A Complex of C9ORF72 and P62 Uses Arginine Methylation to Eliminate Stress Granules by Autophagy

ARTICLE

DOI: 10.1038/s41467-018-05273-7 OPEN A complex of C9ORF72 and p62 uses arginine methylation to eliminate stress granules by autophagy

Maneka Chitiprolu 1,2, Chantal Jagow1,2, Veronique Tremblay2,3, Emma Bondy-Chorney1, Geneviève Paris1, Alexandre Savard1,2, Gareth Palidwor4, Francesca A. Barry2,3, Lorne Zinman5, Julia Keith5, Ekaterina Rogaeva6, Janice Robertson6, Mathieu Lavallée-Adam2,3, John Woulfe7, Jean-François Couture2,3, Jocelyn Côté 1 & Derrick Gibbings1,2 1234567890():,;

Mutations in proteins like FUS which cause Amyotrophic Lateral Sclerosis (ALS) result in the aberrant formation of stress granules while ALS-linked mutations in other proteins impede elimination of stress granules. Repeat expansions in C9ORF72, the major cause of ALS, reduce C9ORF72 levels but how this impacts stress granules is uncertain. Here, we demonstrate that C9ORF72 associates with the autophagy receptor p62 and controls elimination of stress granules by autophagy. This requires p62 to associate via the Tudor protein SMN with proteins, including FUS, that are symmetrically methylated on arginines. Mice lacking p62 accumulate arginine-methylated proteins and alterations in FUS-dependent splicing. Patients with C9ORF72 repeat expansions accumulate symmetric arginine dimethylated proteins which co-localize with p62. This suggests that C9ORF72 initiates a cascade of ALS-linked proteins (C9ORF72, p62, SMN, FUS) to recognize stress granules for degradation by autophagy and hallmarks of a defect in this process are observable in ALS patients.

1 Department of Cellular and Molecular Medicine, University of Ottawa, 451 Smyth Road, Ottawa, Ontario K1H 8M5, Canada. 2 Ottawa Institute of Systems Biology, University of Ottawa, 451 Smyth Road, Ottawa, Ontario K1H 8M5, Canada. 3 Department of Biochemistry, Microbiology and Immunology, University of Ottawa, 451 Smyth Road, Ottawa, Ontario K1H 8M5, Canada. 4 Ottawa Bioinformatics Core Facility, Ottawa Hospital Research Institute, 501 Smyth Road, Ottawa, Ontario K1H 8L6, Canada. 5 Sunnybrook Health Sciences Centre, 2075 Bayview Avenue, Toronto, Ontario M4N 3M5, Canada. 6 Tanz Centre for Research in Neurodegenerative Disease, University of Toronto, Krembil Discovery Tower, 60 Leonard Avenue, Toronto, Ontario M5T 2S8, Canada. 7 Department of Pathology and Laboratory Medicine, University of Ottawa, 501 Smyth Road, Ottawa, Ontario K1H 8L6, Canada. Correspondence and requests for materials should be addressed to D.G. (email: [email protected])

NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7

n autophagy, a phagophore elongates to engulf and then C9ORF72 siRNA (Fig. 1a, Supplementary Fig. 1a, b) confirming Ienclose cytoplasmic contents in an autophagosome. Fusion of the specificity of the immuno-enrichment. C9ORF72 immuno- the autophagosome with lysosomes generates an autophago- precipitated a known C9ORF72 interactor (hnRNPA2B1) lysosome and results in cargo degradation1. Cargoes are fre- (Fig. 1b). Mass spectrometry of three biological replicates of quently recruited for degradation by autophagy in a selective C9ORF72 immunoprecipitates compared to IgG controls identi- manner. Autophagy receptors such as p62 (Sequestosome-1 fied 309 proteins putatively interacting with C9ORF72 (Bayesian (SQSTM1)) bind both specific cargoes and LC3, an integral part False Discovery Rate < 0.05) as assessed by the Significance of the elongating phagophore, thereby directing the selective Analysis of INTeractome (SAINT) algorithm (Supplementary recruitment of cytoplasmic cargoes into autophagosomes2,3.In Fig. 1c, Supplementary Table 1). Stringently limiting candidates most published cases ubiquitination of substrates is required for to those enriched in C9ORF72 immunoprecipitation vs. IgG their recognition by selective autophagy receptors, but in some control in each of the three biological replicates yielded 36 high- cases degradation appears to be independent of ubiquitination4. confidence interactors including multiple previously identified Early studies demonstrated that a large proportion of RNA C9ORF72 interactors (Fig. 1c, Supplementary Table 1). Strikingly, degradation in stress is performed by autophagy5. It is now 8 (22%) of the C9ORF72 interactors during oxidative stress were increasingly clear that cytoplasmic RNA granules, including stress known stress granule components20–22 including two ALS-linked granules, are degraded by autophagy6–8. Stress granules coalesce proteins (hnRNPA2B1 and hnRNPA1, Fig. 1c, Supplementary RNA and RNA-binding proteins in large cytoplasmic clusters Table 1). FUS and SMN, two other ALS-linked proteins that within minutes of stresses such as oxidative stress9,10. When localize to stress granules, also immunoprecipitated with stressors are removed, many stress granules disassemble, but a C9ORF72-myc (Fig. 1c, Supplementary Fig. 1d, e). Putative significant proportion relies on autophagy for their elimination7. C9ORF72 interactors were enriched in RNA binding and RNA The rapid concentration of select RNA-binding proteins con- splicing GO terms (Supplementary Table 2). Upon exposure to trolling splicing and translation in stress granules is postulated to oxidative stress induced by arsenite, multiple RNA-binding pro- re-shape the post-transcriptional landscape to rapidly tailor cel- teins including PABP, TIAR, FMRP, FUS, SMN and DDX3 were lular responses to stress9,10. Interestingly, several mutations recruited into stress granules (Supplementary Fig. 1d, e). Con- genetically linked to ALS impair nuclear localization of RNA- sistent with putative interactions of C9ORF72 with stress gran- binding proteins like FUS and TDP-43 causing them to accu- ules, C9ORF72 was recruited into 93% of FMRP + stress granules mulate in cytoplasmic stress granules and insoluble inclusions in (Fig. 1d, Supplementary Fig. 1f, g). This suggests that C9ORF72 patient neurons11. Importantly, evidence suggests that pathology has a role in stress granule biology. in ALS is caused not only by loss of the splicing capacity in the Intriguingly, a third ALS-linked protein that is a high- nucleus of proteins like FUS, but also by a toxic gain of function confidence candidate interactor with C9ORF72 is p62 (Supple- in the cytoplasm12. This has led to a model in which inap- mentary Table 1). C9ORF72 immunoprecipitated p62, and HA- propriate induction of stress granules or their derivatives may p62 reciprocally immunoprecipitated C9ORF72 both in the play a role in ALS pathology13. absence and presence of oxidative stress (Fig. 1a, b, e). Proximity Intriguingly, mutations in valosin-containing protein (VCP), ligation assays (PLA) confirmed that in intact cells p62 and which are an inherited cause of ALS, control elimination of stress C9ORF72 closely associate. While no PLA signal was detected granules by autophagy7. P62 localizes to many types of inclusions with control non-specific antibodies (Fig. 1f), or between in ALS patients. We previously demonstrated that p62, which is C9ORF72 and another autophagy receptor NBR1 (Fig. 1g), genetically linked to ALS14, degrades stress granule-like cyto- prominent PLA signals were obtained between p62 and a known plasmic aggregates by selective autophagy6. This suggests that interactor (LC3) as well as between C9ORF72 and p62 (Fig. 1f, h). either instigating formation of inclusions related to stress granules PLA signal between C9ORF72 and p62 was lost upon knockdown or impeding their clearance by selective autophagy may have a of C9ORF72 (Fig. 1h, i). Finally, more recombinant p62 was role in ALS. pulled down with recombinant purified GST-C9ORF72 than with Repeat expansions in an intron of C9ORF72 are the most beads containing GST alone (Fig. 1j). Together, this suggests that common genetic cause of ALS15,16. Repetitive RNAs and dipeptide C9ORF72 and p62 are part of a physically-associated complex. repeat proteins are produced by transcription and translation of We investigated whether C9ORF72 and p62 have a shared role these repeat expansions and overexpression of these can induce in stress granule biology. We immunoprecipitated HA-p62 ALS-like pathology17. C9ORF72 repeat expansions also cause (Fig. 1k). p62 interactors (vs. IgG control immunoprecipitates) decreased levels of C9ORF72 mRNA and protein, suggesting that with and without oxidative stress were identified by mass alongside repeat-induced pathology certain aspects of ALS spectrometry analysis using SAINT (Supplementary Table 3, pathology could be caused by loss of C9ORF72 function18,19. Fig. 1l). In addition to immunoprecipitating LC3 and ubiquiti- Here, we show that symmetric arginine methylation of stress nated proteins as anticipated, p62 immunoprecipitated candidate granules by Protein Arginine Methyltransferase 5 (PRMT5) is interactor DDX3, but not EWS another RNA-binding protein required for a complex of C9ORF72 and p62 to associate with (Fig. 1m). Candidate p62 interactors overlapped with putative stress granules and eliminate them by autophagy. Additionally, p62 interactors from other studies curated in iRefWeb23 by 26.3% p62−/− mice and patients with C9ORF72 repeat expansions show (20/76 proteins) in untreated cells and 26.1% (12/46 proteins) in signs of defects in this process. arsenite-treated cells compared to random overlap of 2.9% (431 putative interactors/15,000 proteins in the proteome). Like C9ORF72, candidate p62 interactors were strongly enriched in Results RNA splicing functions and 29% of p62 interacting proteins were C9ORF72 associates with stress granule proteins and p62.To RNA-binding proteins (Supplementary Table 4). Seven (15%) of identify new functions of C9ORF72 during oxidative stress, we the candidate p62 interactors were known stress granule immunoprecipitated endogenous C9ORF72 from cells exposed to components (Fig. 1l, Supplementary Table 5). p62 also interacted arsenite-induced oxidative stress (+As) and untreated counter- with SMN and other RNA-binding proteins including FUS parts (−As, Fig. 1a). C9ORF72 was detected as expected at 50 (Fig. 1n), which localize to stress granules (Supplementary kDa with a secondary antibody specific for IgG light chain Fig. 1d, e). p62 interaction with FUS was independent of RNA (Fig. 1a). This band was diminished in cells treated with a (Supplementary Fig. 1h, i). Strikingly, p62 and C9ORF72

2 NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 ARTICLE interactomes upon oxidative stress signiﬁcantly overlapped (p < numbers of stress granules (30 min arsenite) as control cells 0.0001, Chi-test) including RNA-binding proteins known to (Fig. 2a–d). In contrast, during recovery from stress (1 h recovery localize to stress granules (Supplementary Table 6). This strongly after 30 min arsenite), 88 or 84% of cells, respectively, treated with suggests that a complex of C9ORF72 and the autophagy receptor siRNA targeting p62 (Fig. 2a, b) or C9ORF72 (Fig. 2c, d) failed p62 (Fig. 1b, e–j, Supplementary Table 1) associates with stress to eliminate stress granules, closely mimicking the effect granule proteins in cells exposed to oxidative stress (Fig. 1b–d, of inhibiting autophagy with ATG5 siRNA (Supplementary k–n). Fig. 2a, b). p62 knockdown had no effect on stress granule numbers in ATG5−/− cells suggesting that p62 operates in a C9ORF72 and p62 help eliminate stress granules by autophagy. ATG5-dependent autophagy pathway to clear stress granules To test if C9ORF72 and p62 control elimination of stress gran- (Supplementary Fig. 2c). In contrast, depletion of other autop- ules, p62 or C9ORF72 were depleted with siRNA. Depletion of hagy receptors including NBR1 and OPTN did not affect clear- C9ORF72 had minimal effect on p62 levels (Supplementary ance of stress granules (Supplementary Fig. 2d). Consistent with Fig. 1b). Cells depleted of p62 or C9ORF72 formed similar degradation of cytoplasmic stress granules by p62-dependent

IP ab–Arsenite +Arsenite –Arsenite +Arsenite Input IgG C9 IP Input IgG C9 IP Input IgG C9 IP Input IgG C9 IP SG Proteins InputBeadsC9-myc siRNA: HNRNP 36 Ctl Ctl Ctl Ctl Ctl Ctl C9 C9 C9 C9 C9 C9 A2B1 8 myc C9ORF72 p62 (Proteintech) p62 IgG HC SMN C9ORF72 2 Interactors 3 FUS ALS-linked d

100

80 e

60 Input IgG IP HA(p62) IP –As +As –As +As –As +As +Arsenite 40 C9ORF72 20 HA (p62) 0 DAPI C9ORF72 (Santa Cruz) FMRP Merge % Stress granules with C9ORF72

f g +Arsenite hi+Arsenite 45 Ctl siRNA 1.62E-03

40 C9ORF72 siRNA **

30 Ctl siRNA

C9ORF72-NBR1 25 p62-C9ORF72 Proximity ligation assay 20 (+Arsenite) 15 0.0293 10 * Proximity ligation assay (PLA)

NBR1 5 C9ORF72-p62 PLA signal counts/cell, C9ORF72 siRNA

C9ORF72-p62 proximity ligation assay (PLA) 0 +Baf A1, p62-LC3 PLA IgG-IgG PLA Immunofluorescence p62-C9ORF72 Nucleus Cytoplasm

j k mn Input IgG HA(p62) IP Input IgG HA(p62) IP –Arsenite +Arsenite – + – + – + As Input IgG p62 IP Input IgG p62 IP –As +As –As +As –As +As HA (p62) LC3-I LC3-I GST GST-C9ORF72 LC3-II LC3-II p62 l p62 GFP (DDX3)

GST-C9ORF72 SMN EWS

46 SG HA (p62) Input IgG HA(p62) IP Proteins IP –As +As –As +As –As +As Input IgG HA-p62 GST 7 FUS p62 Interactors LC3-I LC3-II 75 Ub

63 HA (p62) HA (p62) IgG HC

NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 autophagy, cytoplasmic fractions, but not nuclear fractions, fluorescence found no images with greater co-localization than accumulated the stress granule protein FUS in cells treated with input images of p62 and TIAR). This closely resembles the type siRNA targeting p62, or genes required for autophagy (ATG5 and and amount of localization of autophagy receptors or LC3 to ATG7, Fig. 2e, f, Supplementary 2b). other autophagy substrates27–29 including stress granules7. This If p62 directly affects clearance of stress granules (Fig. 2a, b, amount of co-localization is also consistent with the fact that not Supplementary 2a–c); it may localize to stress granules. P62 all stress granules are degraded by autophagy7, p62 is involved in associated with the stress granule proteins FUS, SMN, and DDX3 degradation of other substrates, and autophagic degradation is by immunoprecipitation (Fig. 1m, n). PLA confirmed that p62 rapid (~10 min). and FUS associate in intact cells (Fig. 2g). To test whether p62 Since C9ORF72 and p62 form a complex (Fig. 1b, e–j) and physically associates with stress granules we used electron control elimination of stress granules (Fig. 2a–d), they may dock microscopy. As expected, p62-labeled mitochondria24,25, as well on stress granules. Endogenous C9ORF72 and p62 co-localized as endolysosomal structures with heterogenous content resem- both in untreated and arsenite-treated cells (Fig. 2n, o). PLA bling autophagolysosomes (Supplementary Fig. 2e). In arsenite- detected an association of C9ORF72 and p62 on the border of treated cells, stress granules were observed, as previously some stress granules (Supplementary Fig. 2i). Together, while described26, as electron-dense structures labeled with the stress C9ORF72 may also have effects on stress granules independent of granule markers FUS and FMRP (Fig. 2h). p62 frequently p62, this suggests that C9ORF72 and p62 form a complex that clustered on the periphery of stress granules (Fig. 2i). FUS and docks on stress granules to control their elimination by FMRP were also observed on similar electron-dense structures autophagy either as large structures or portions of stress granules inside autophagolysosome-like structures (Fig. 2h), suggesting (Figs. 1, 2, Supplementary Figs. 1, 2). their degradation by autophagy. In agreement, p62 and FUS were infrequently found in close proximity in the cytoplasm (Fig. 2k) but 28% of structures co-labeled by p62 and FUS were stress PRMT5 helps eliminate stress granules by autophagy.We granules and another 50% were inside autophagolysosomes inquired how a C9ORF72-p62 complex eliminates stress granules (Fig. 2j, k). In agreement with the role of autophagy in by autophagy. Canonically, p62 recruits ubiquitinated cargoes for engulfment of stress granules, p62-labeled phagophore-like degradation by autophagy3. Ubiquitin was not enriched in stress structures also associated with stress granules (Fig. 2j), stress granules despite forming punctae in cells treated with inhibitors granules labeled with FUS and p62 were observed within double- of autophagic degradation and co-localizing with OPTN, a membrane autophagosomes (Fig. 2j) and stress granule proteins ubiquitin-binding autophagy receptor (Fig. 3a, Supplementary TIAR and SMN accumulated in cells depleted of ATG5 Fig. 3a, b). This suggested that p62 may be recruited to stress (Supplementary Fig. 2f). Together, this suggests that either entire granules by a non-canonical signal. Interestingly, arginine- stress granules or fragments of stress granules are degraded by dimethylated proteins, particularly symmetrically dimethylated p62-dependent autophagy and this is required for complete proteins (SYM), were strongly enriched in stress granules (Fig. 3b, elimination of stress granules from cells (Fig. 2a–f, Supplementary c). Asymmetrically dimethylated proteins (ASYM) were also Fig. 2a–c). found in stress granules (Fig. 3b, c). Subcellular fractions were Intriguingly, by electron microscopy p62 was frequently prepared according to established protocols30 to enrich autop- observed on the periphery of stress granules (Fig. 2i, j), as hagosomes and autophagolysosomes. Accordingly, these fractions required to remain accessible to LC3 and direct their degradation were enriched in LC3-II (vs. LC3-I) compared to the LC3-II/LC3- by autophagy. In agreement with this, 3D reconstructions of I ratio in the corresponding total cell lysates (Fig. 3d). Comparing confocal microscopy images demonstrated that p62 docked on the ratio of proteins in the autophagolysosome fractions to their stress granules and LC3 coated p62 on these stress granules level in the corresponding cell lysate indicates the propensity of (Fig. 2l, Supplementary Fig. 2g, h). By immunofluorescence, p62 proteins to be degraded by autophagy. Autophagolysosome docked on 21% of stress granules (Fig. 2m, Supplementary 2g, h), fractions contained the canonical autophagy substrate p62, stress closely paralleling the percent of stress granules (28%) containing granule proteins degraded by autophagy such as FUS (Fig. 2e–m), p62 by electron microscopy (Fig. 2j, k). Costes’ test demonstrated and SYM at similar autophagosome fraction/cell ratios (Fig. 3d), that co-localization of p62 with TIAR in the cytoplasm is not suggesting that proteins with symmetrically dimethylated argi- random (in each of 88 cells, 100 iterations of randomized nines are degraded at a rate comparable to p62 and stress

Fig. 1 C9ORF72 associates with ALS-linked stress granule proteins including p62. a Western blot of endogenous C9ORF72 immunoprecipitates compared to immunoprecipitation with non-specific IgG from cell lysates (input) transfected with control or C9ORF72 siRNA. Secondary antibody specific for IgG light chain was used. b Western blot of endogenous C9ORF72 immunoprecipitates with and without arsenite treatment. c Left, Venn-diagram depicting the number of proteins in the putative C9ORF72 interactome that localize to stress granules (SGs, blue) or contain mutations linked to ALS (olive); Right, western blot of immunoprecipitation of C9ORF72-myc for p62 and SG proteins (FUS, SMN). d Immunofluorescence of endogenous C9ORF72 and SGs (FMRP) after 30 min incubation with Sodium Arsenite, (0.5 mM); Right, quantification of the percentage of SGs (FMRP+) co-localizing with foci of C9ORF72 (quantified 3916 C9ORF72 punctae and 3331 SGs). e Western blot of HA-p62 immunoprecipitates with and without arsenite treatment. f PLA co-labeling with species controlled IgG (top, negative control) and known interactors p62 and LC3 (bottom, positive control); p62-LC3 association was tested in cells treated with Bafilomycin A1 (Baf A1). g Top, proximity ligation assay (PLA) for C9ORF72 and NBR1 in arsenite-treated cells; Bottom, immunofluorescence for NBR1 showing antibody does label cells. h PLA for endogenous p62 and C9ORF72 in cells treated with arsenite and transfected with control (Ctl) or C9ORF72 (C9) siRNA. i Quantification of PLA signals for C9ORF72 and p62 in the cytoplasm and nucleus in identical conditions to (h)(n = 3, mean ± SD, Student’s t-test). j Western blot of recombinant p62 and GST-C9ORF72 from in vitro pulldown assay. k Representative immunoprecipitation of HA-p62 used for identification of putative p62 interacting proteins using LC-MS/MS. l Venn-diagram depicting number of proteins in predicted p62 interactome that localize to SGs (blue). m Western blot of immunoprecipitates of HA-p62 detecting known interactors (LC3, ubiquitinated proteins), candidate interactors identified by LC-MS/MS (GFP-DDX3) and control RNA-binding proteins that do not associate with p62 (EWS) in cells treated or not with arsenite. n Western blot of immunoprecipitations of endogenous p62 (top) and HA-p62 (bottom) for other SG proteins (FUS, SMN). HeLa cells were used in all experiments. Scale bar = 10 µm. Molecular weight markers for blots, where not shown, are included in the uncropped versions

4 NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 ARTICLE granules. Additionally, SYM accumulated upon depletion of We tested whether symmetric arginine dimethylation is ATG7 and ATG12 (Supplementary Fig. 3c). p62 and LC3 also required for elimination of stress granules. The methylation immunoprecipitated with proteins containing symmetrically inhibitor, methylthioadenosine (MTA) (Supplementary Fig. 3d) dimethylated arginines (Fig. 3e), alongside proteins known to alone did not induce stress granules and had minor effects on cell contain (Sm) or bind to (SMN) this post-translational mod- survival (Supplementary Fig. 3e). In cells treated with arsenite, iﬁcation (Fig. 3e). This suggests that proteins with symmetrically MTA did not affect stress granule formation but prevented their dimethylated arginines are enriched in stress granules, associate elimination (Fig. 3f, g). Depletion of the symmetric dimethylation with p62 and are degraded by autophagy. enzyme PRMT5 (Fig. 3h, Supplementary Fig. 3f, h–i) did not

0.03 a Ctl siRNA c Ctl siRNA e * Ctl siRNA 0.01 p62 siRNA C9ORF72 siRNA * p62 siRNA 2.24E-07 4.07E-10 3.0 100 ***100 *** ATG5 siRNA 2.5 0.02 ATG7 siRNA 80 80 * 2.0 60 60 1.5 40 40 1.0

20 20 Relative protein level 0.5

% cells with stress granules 0 % cells with stress granules 0 30’ arsenite 30’ arsenite 30’ arsenite 30’ arsenite 0 +1 h recovery +1 h recovery Cytoplasmic-FUS Nuclear-FUS

bdfctl siRNAp62 siRNA ctl siRNA C9ORF72 siRNA TIAR TIAR FMRP FMRP Cytoplasmic Nuclear

siRNA: Ctl p62 ATG5 ATG7 Ctl p62 ATG5 ATG7 30 min As 30 min As FUS 75

TIAR TIAR FMRP FMRP Histone H3K27 17 43 GAPDH 34 30 min As, 1h recovery 30 min As, 1 h recovery g h i Ctl siRNA FUS siRNA FMRP FMRP p62

Stress granule Autophagolysosome Stress granule

FUS FUS p62 +Arsenite (+As) +Arsenite (+As) +Arsenite (+As) Stress granule Autophagolysosome

p62-FUS p62-FUS Stress granule p62-FUS proximity ligation assay (PLA)

j FUS (18 nm) p62 (12 nm) FUS (18 nm) FUS (18 nm) FUS (18 nm) p62 (12 nm) p62 (12 nm)

Stress granule FUS

A utophagosome FUS

FUS (18 nm) FUS (18 nm) Putative A utophagosome p62 (12 nm) p62 (12 nm) FUS phagophore +Arsenite (+As) FUS p62 p62 Stress granule Stress granule

Autophagolysosome

klCytoplasm Electron dense (stress granule) 0.3 µm 0.5 µm 0.4 µm Autophagolysosome 80

20 p62 Percent of total gold labeling 0 p62 p62 FMRP p62 FUS p62 only only and FUS FUS FMRP GFP-LC3 mn o C9ORF72 (Santa Cruz) p62 Merge 30 30

25 25

20 20 – Arsenite (–As) 15 15 C9ORF72 (Santa Cruz) p62 Merge 10 10

5 5

0 0 +Arsenite (+As) % of structures with overlapping punctae p62 punctae SGs with –Arsenite +Arsenite with SGs p62 punctae % C9ORF72 punctae overlapping with p62

NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 affect autophagosome formation or p62 levels (Supplementary The Tudor protein SMN mediates association of FUS and p62. Fig. 3f, g). PRMT5 depletion abrogated symmetric arginine We investigated how p62 associates with arginine-methylated dimethylation in stress granules (Fig. 3h) and prevented their FUS. Tudor domains bind dimethylated arginines32. Survival elimination during recovery from stress (Fig. 3i, j). A PRMT5 motor neuron protein (SMN) is a Tudor protein that pre- inhibitor (EPZ015666) had similar effects (Supplementary ferentially binds symmetrically methylated arginines33, frequently Fig. 3j-m). Cumulatively, this suggests that like C9ORF72 and co-localizes with or docks on stress granules like p62 (Fig. 4g)34, p62, symmetric arginine dimethylation of stress granule proteins and associates with FUS35 (Fig. 4c, Supplementary Fig. 4a). Like by PRMT5 is required for their elimination. Knockdown of FUS, SMN immunoprecipitated with p62 in the absence or pre- C9ORF72, p62, or PRMT5 decreased recruitment of LC3 to the sence of oxidative stress (Figs. 1n, 4h). This suggests that SMN stress granule marker FMRP as measured by PLA (Fig. 3k, may link FUS to p62. Supporting this mechanism, association of Supplementary 3n) independent of any change in LC3 or FMRP FUS with p62 was reduced when SMN was depleted with siRNA levels (Fig. 3l). With evidence that p62 is epistatic to ATG5 in (Fig. 4i, Supplementary Fig. 4b). eliminating stress granules (Supplementary Fig. 2a–c), and the We sought to investigate binding of p62 to symmetrically presence of stress granule markers with p62 in autophagosomes methylated arginines and SMN in more detail using recombinant and autophagolysosomes by density gradient (Fig. 3d), confocal proteins. Recombinant SMN was pulled down with recombinant microscopy (Fig. 2l), and electron microscopy (Fig. 2h–k), this GST-p62, but not GST alone (Fig. 4j). In agreement with direct strongly suggests that a complex of C9ORF72 and p62 (Fig. 1) binding of recombinant p62 and SMN, p62 binding to SMN in requires PRMT5 to eliminate stress granules by autophagy. cells did not require the p62 ubiquitin-binding domain (UBA) or its LC3 interacting region (LIR) (Supplementary Fig. 4c). In subsequent assays, peptides containing symmetrically dimethy- Methylation of FUS by PRMT5 is required for binding to p62. lated peptides were biotin-labeled and used for pulldowns Candidate interactors of C9ORF72 and p62 included several compared to no peptide or identical peptides lacking symmetric proteins containing dimethylated arginines by mass spectrometry, dimethylation. As expected33, SMN exhibited more binding to including FUS (Supplementary Table 6). We asked whether FUS peptides containing symmetric arginine dimethylation (Fig. 4k). could be symmetrically dimethylated by PRMT5. PRMT1 is 31 p62 exhibited some background binding to beads likely due to its known to asymmetrically dimethylate arginines in FUS , how- tendency to oligomerize and aggregate, but in the presence of ever FUS is not known to be symmetrically dimethylated, or SMN, p62 bound more to beads containing symmetrically dimethylated by PRMT5. FUS was detected in immunoprecipi- dimethylated peptides (Fig. 4k). This demonstrates that p62 can tates of proteins with symmetrically dimethylated arginines bind the Tudor protein SMN (Fig. 4j) and requires SMN to bind (Fig. 4a), and a protein at the mass of FUS was detected with ﬁ to peptides containing symmetric dimethylation (Fig. 4k, l). antibodies speci c for symmetrically dimethylated arginines in Replicating the phenotype of p62 and PRMT5 knockdown FUS immunoprecipitates (Fig. 4b). PRMT5 was detected in FUS (Figs. 2a, b, 3i, j), depletion of SMN did not impede formation immunoprecipitates along with SMN which binds SYM (Fig. 4c), of stress granules but blocked their elimination during recovery suggesting that PRMT5 may symmetrically dimethylate FUS. In from stress (Fig. 4m, Supplementary 4d). This suggests that p62 agreement, symmetric dimethylation of arginine residues on FUS uses SMN to recognize arginine dimethylated proteins including was reduced upon PRMT5 depletion (Fig. 4b) and PRMT5 FUS in stress granules for clearance by autophagy. methylated GST-FUS, but not GST in an in vitro radiolabeling assay (Fig. 4d). This suggests that PRMT5 symmetrically dime- thylates FUS in vitro and in vivo. PRMT5 methylates FUS on R218 for autophagic degradation. Intriguingly, FUS was only detected in p62 immunoprecipitates PRMT5-mediated symmetric dimethylation promotes FUS asso- by mass spectrometry in its arginine dimethylated form ciation with p62 (Fig. 4a-f). FUS was previously observed to (Supplementary Tables 5, 6). This suggests that dimethylation associate with symmetrically arginine-dimethylated proteins36, of FUS may be required for its binding to p62. In agreement, FUS but whether and where it was directly symmetrically methylated immunoprecipitated with p62 and this association was strongly remained unclear. FUS detected by mass spectrometry in both reduced in cells treated with the methylation inhibitor MTA or p62 and C9ORF72 immunoprecipitates was dimethylated exclu- depleted of PRMT5 (Fig. 4e, f). sively on a peptide encompassing two arginines (R216, R218,

Fig. 2 C9ORF72 and p62 promote stress granule clearance by autophagy. a Percentage of cells containing stress granules in cells treated with control or p62 siRNA after 30 min arsenite, or one hour after removal of arsenite (n = 3, mean ± SD, Student’s t-test). b Representative images of stress granules used in (a). c Percentage of cells containing stress granules in cells treated with control or C9ORF72 siRNA after 30 min arsenite, or one hour after removal of arsenite (n = 3, mean ± SD, Student’s t-test). d Representative images of stress granules used in (c). e Quantification of FUS levels in cytoplasmic and nuclear extracts of cells treated with siRNA targeting p62, ATG5, ATG7, or control siRNA (n = 5, mean ± SD, Student’s t-test). f Representative western blot of (e). g Proximity ligation assay for p62 and FUS in arsenite-treated cells transfected with control siRNA or siRNA targeting FUS. Scale bar = 10 µm. h Immuno-electron micrographs of FUS and FMRP in cells treated with arsenite; arrows highlight antibody labeling within indicated stress granule-like structures. i Immuno-electron micrographs of p62 in cells treated with arsenite. j Immuno-electron micrographs of p62 (12 nm gold) and FUS (18 nm gold) in cells treated with arsenite; arrows highlight co-labeled structures. Scale bar = 100 nm. k Quantification of 46 randomly selected immuno-electron micrographs co-labeled with p62 and FUS as in (j). Each electron-dense stress granule-like structure, autophagolysosome-like organelle or 50 nm radius area of cytoplasm with label was quantified as being labeled with p62 alone, FUS alone, or both p62 and FUS. l 3D reconstruction of Z-stack confocal images of cells. Images are zoomed-in on stress granules labeled with the indicated antibodies or fluorescent proteins (GFP-LC3). m Quantification of the percentage of p62 punctae co-labeled with stress granule markers and the percentage of stress granules with co-localized or docked p62 (quantified 12102 p62 punctae and 12627 stress granules, n = 6, mean ± SD, Student’s t-test). n Confocal image of endogenous C9ORF72 and p62 in cells treated with or without arsenite. Scale bar = 10 µm. o Quantification of the percentage of C9ORF72 punctae overlapping with p62 in cells treated with or without arsenite (quantified C9ORF72 punctae without arsenite: 2688, with arsenite: 3778 punctae; p62 punctae without arsenite: 1805, with arsenite: 2942, n = 2, mean ± SD, Student’s t-test). HeLa cells were used in all experiments

6 NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 ARTICLE

Supplementary Table 6). Other dimethylation sites were identi- To confirm this, R218K mutation of FUS strongly decreased its fied; however methylation at R218 was the only site consistently symmetric arginine dimethylation detected by western blot identified in assays including p62 and C9ORF72 interactomes, (Fig. 5a, Supplementary Fig. 5c). This suggests that R218 is a increasing in methylation assays with PRMT5 (Fig. 4d, Supple- principal site of PRMT5-dependent symmetric dimethylation on mentary Table 7), and decreasing in cells expressing FUS. PRMT5 shRNA (Supplementary Fig. 5a). In some cases, FUS PRMT5-dependent methylation of FUS promotes its binding peptides were fragmented between R216 and R218, and this to SMN and p62 (Fig. 4f–l). FUS R218K mutants and others were confirmed FUS is dimethylated on R218 (Supplementary Fig. 5b). expressed at similar or slightly higher levels in cells (Fig. 5b,

abc

DDX3 Ub Merge (+As) +Arsenite SYM FMRP Merge

–Arsenite (–As) –Arsenite SYM

–Arsenite (–As) –Arsenite

DDX3 Ub Merge ASYM FMRP Merge

ASYM (+As) +Arsenite

+Arsenite (+As) +Arsenite

de Input IgG SYM IP –As +As –As +As –As +As 72 HA (p62) 56

Atg fraction #1 fraction Atg #2 fraction Atg Total cell lysate cell Total 170 43 130 SMN 95 34

72 lysate cell Total #1 fraction Atg #2 fraction Atg

Total cell lysate cell Total #1 fraction Atg #2 fraction Atg 56 43 72 SMN +Arsenite p62 43 56 34 Input IgG SYM IP 34 SYM 43 TIAR SMN LC3-I 17 26 Sm (Y12) LC3-II LC3-I 17 LC3-I 95 LC3-II FUS LC3-II 72 17

f giControl Methylation inhibitor (MTA) sh(Luciferase) DMSO sh(PRMT5) MTA 3.3E-05 1.41E-08 100 100 *** *** 80 80 30 min As

60 TIAR TIAR 60

40 40

20 20 % cells with stress granules % cells with stress granules 0 0 ′ ′ 30 arsenite 30 arsenite 30′ arsenite 30′ arsenite +1 h recovery TIAR TIAR

30 min As, 1 h recovery +1 h recovery

hjCtl shRNA (luciferase) PRMT5 shRNA FMRP FMRP 30 min As

SYM FMRP Merge Ctl shRNA (luciferase) FMRP FMRP + Arsenite

PRMT5 shRNA SYM FMRP Merge 30 min As, 1 h recovery

klLC3-FMRP proximity ligation assay (PLA), +Arsenite Ctl siRNA C9ORF72 siRNA

10 0.0046 siRNA: Ctl C9 p62 PRMT5 0.041 95 ** * 72 FMRP 0.0003 Tubulin p62 siRNA PRMT5 siRNA *** 48 (+Arsenite) 17 LC3-I 11 LC3-II

LC3-FMRP PLA signal counts / cell 0 Ctl C9ORF72 p62 PRMT5

NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7

Supplementary Fig. 5d). In agreement with R218 of FUS being the impaired compared to their wild-type counterparts (Fig. 5f). Both principal site of symmetric dimethylation by PRMT5, the FUS and SMN immunoprecipitated with SYM in mouse cortex, immunoprecipitation of p62 and SMN with FUS was reduced cerebellum, and spinal cord (Supplementary Fig. 5i), suggesting by mutation of R218 (Fig. 5b). In agreement, similar amounts of that FUS is methylated by PRMT5 in relevant mouse tissues. SYM FUS R218K and other FUS mutants localized to the cytoplasm as and FUS accumulated in cerebellum, but not the hippocampus of wild-type FUS during oxidative stress (Fig. 5d, Supplementary p62−/− mice (Fig. 5g). While we cannot exclude that p62 pro- Fig. 5e), but in PLA p62 association with FUS was ablated by the motes symmetric arginine dimethylation, this data suggests that R218K mutation (Fig. 5c). Nonetheless, depletion of FUS with p62 controls turnover of stress granules in neurons and proteins siRNA did not alter formation or elimination of stress granules containing symmetrically dimethylated arginines in some regions (Supplementary Fig. 5f), suggesting that p62 recognizes multiple of the mouse brain. proteins methylated by PRMT5 to promote elimination of stress granules. If p62 can also eliminate symmetrically methylated p62−/− mice exhibit FUS-dependent mRNA splicing defects. proteins individually or in small clusters then mutating methyla- P62 may impact the amount and nuclear-cytoplasmic distribution tion sites should block their turnover by autophagy. Wild-type of FUS and other splicing regulators that are symmetrically fi − − FUS signi cantly accumulated when autophagic degradation was dimethylated resulting in alternative splicing. In brain of p62 / fi blocked with Ba lomycin A1 whereas R218K mutant was mice37, while no mRNAs besides p62 exhibited changes in unperturbed demonstrating that it is no longer targeted for expression, 1386 mRNAs exhibited signatures of alternative degradation by autophagy (Fig. 5e, Supplementary Fig. 5g). This splicing. A predicted p62-dependent splicing event in Zcchc6 strongly suggests that symmetric arginine dimethylation of FUS (exon 27 inclusion) was impaired in p62 knockdown MN-1 cells at R218 by PRMT5 is required for its association with p62 and and in several brain regions of p62−/− mice (Fig. 5h, i, Supple- clearance by autophagy (Figs. 3i-k, 4, 5a-e), while methylation of mentary Fig. 5j–l). Zcchc6 uridylates the 3′ end of select miRNA multiple proteins in stress granules is likely required for their and mRNAs including let-738 whereby it may impact formation elimination by autophagy. of neuromuscular junctions39 and ALS-associated phenotypes40. Like FUS R218K, the association of FUS R216C and R521H but Exon 27 inclusion of Zcchc6 mRNA was increased when FUS was not R521G with p62 was strongly decreased, despite accumulating depleted with siRNA, reversed when wild-type FUS, but not FUS similarly in the cytoplasm and being expressed at similar levels mutants (R218K, R216C, R521G, R521H) were expressed and – (Fig. 5b d, Supplementary Fig. 5e). The association of FUS with further enhanced by blocking autophagic degradation of FUS p62 could be impaired at multiple levels. For example, PRMT5 or with Bafilomycin (Fig. 5j, l, Supplementary Fig. 5m–o). The SMN may no longer be able to access R218, or the FUS-SMN decreased inclusion of Zcchc6 exon 27 induced by loss of p62 complex could be otherwise altered to impair p62 binding. (Fig. 5h, i) was prevented by depletion of FUS (Fig. 5k), Symmetric dimethylation of FUS was not decreased in R216C or demonstrating that p62 effects on splicing of Zcchc6 require FUS. R521H mutants (Supplementary Fig. 5h). While FUS R216C Overall, this strongly suggests that p62 and FUS regulate shared association with SMN was not impaired, FUS R521H mutant alternative splicing events that depend in part on methylation of associated poorly with SMN (Supplementary Fig. 5h), suggesting R218 on FUS and its ability to associate with p62 and be degraded that the C-terminus of FUS may be required with R218 for by autophagy (Fig. 5a–l, Supplementary Fig. 5a–o). optimal interaction with SMN. FUS R216C was also unperturbed More broadly, mRNAs putatively mis-spliced in p62−/− brains upon inhibiting autophagic degradation (Fig. 5e). Together, this included a highly significant proportion of mRNAs (97 of 862, suggests that multiple mutations in FUS have an impaired p = 1.57 × 10−7) that bound to FUS and were mis-spliced in FUS- association with p62 caused by defects at multiple levels in the deficient brains41 (Fig. 5m, Supplementary Fig. 5p). Furthermore, pathway elucidated here and this can prevent p62-dependent mRNAs mis-spliced in p62−/− brain were significantly enriched in autophagy. a FUS-binding motif41 (Fig. 5n). This suggests that p62 and FUS regulate alternative splicing of a large shared pool of transcripts. p62−/− mice accumulate symmetrically methylated proteins. We assessed whether p62 may affect elimination of stress gran- Patient inclusions contain arginine-methylated proteins. ules in cell types relevant to ALS. Elimination of arsenite-induced C9ORF72 associates with p62 and moderately lowering the levels stress granules in motor neurons derived from p62−/− mice was of C9ORF72 prevents clearance of stress granules (Figs. 1b, e–j,

Fig. 3 Symmetric arginine methylation by PRMT5 is required for stress granule clearance by autophagy. a Immunofluorescent microscopy of DDX3 and ubiquitin (Ub) in cells treated with or without arsenite. b Immunofluorescent microscopy of cells labeled with antibodies for symmetric (SYM) or asymmetric (ASYM) dimethylated arginines. c Immunofluorescent microscopy of cells labeled with antibodies for FMRP and proteins containing symmetric (SYM) or asymmetric (ASYM) arginines after treatment with arsenite. d Western blot of total cell lysate and fractions enriched in autophagosomes (Atg fraction#1) and autophagolysosomes (Atg fraction#2) for known autophagy substrates (LC3-II, p62), stress granule proteins (FUS, SMN, TIAR), and symmetrically dimethylated proteins (SYM). e Western blot of immunoprecipitations of proteins containing symmetrically dimethylated arginines from cells expressing HA-p62 (top) or not (bottom); cells were treated with arsenite (+As) or not (−As). f Percentage of cells containing stress granules in cells treated with DMSO (control diluent) or MTA (methylthioadenosine 1 mM, 48 h) after 30 min arsenite treatment, or one hour after removal of arsenite (n = 3, mean ± SD, Student’s t-test). g Representative images of stress granules labeled with TIAR for (f). h Immunofluorescent images of cells labeled with antibodies specific for symmetrically dimethylated arginines (SYM) or FMRP (stress granule marker) in cells treated with arsenite and expressing either control shRNA or PRMT5 silencing shRNA. i Percentage of cells containing stress granules in cells treated with control shRNA or PRMT5 silencing shRNA after 30 min arsenite treatment, or one hour after removal of arsenite (n = 5, mean ± SD, Student’s t-test). j Representative images of stress granules labeled with FMRP for (i). k Left, PLA for endogenous LC3 and FMRP (in SGs) in cells treated with arsenite and transfected with control (Ctl), C9ORF72 (C9), p62, or PRMT5 siRNA; Right, quantification of proximity ligation assay signals (PLA) for LC3 and FMRP in cells under identical conditions to (k)(n = 3, mean ± SD, Student’s t-test). l Western blot control for PLA in (k) in cells treated with control, C9ORF72, p62, or PRMT5 targeting siRNA. HeLa cells were used in all experiments. Scale bar = 10 µm

8 NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 ARTICLE

a b d In vitro methylation Western blot Input IgG SYM IP Input IgG FUS IP –As +As–As +As –As +As PRMT5 + MEP50 PRMT5 + MEP50 –Arsenite +Arsenite –Arsenite+Arsenite –Arsenite +Arsenite 72 FUS SMN 34 GST GST FUS GST GST FUS 245 25 Sm (Y12) 180 135 100 100 75 GST-FUS 75 GST-FUS SYM 63 63

c sh(Luciferase) sh(PRMT5) sh(Luciferase) sh(PRMT5) sh(Luciferase) sh(PRMT5) sh(Luciferase) sh(PRMT5) sh(Luciferase) sh(PRMT5) sh(Luciferase) sh(PRMT5) blot Input IgG FUS IP 95 48 48 72 FUS GST –As +As–As +As –As +As 55 IgG 35 35 PRMT5 43 Heavy chain 34 26 IgG 25 25 GST SMN Light chain 17 17 FUS 72 FUS PRMT5

f Input IgG HA(p62) IP g

GFP-SMN TIAR Merge e Input IgG HA(p62) IP sh(Luciferase) sh(PRMT5) sh(Luciferase) sh(PRMT5) sh(Luciferase) sh(PRMT5) Ctl MTA Ctl Ctl MTA

72 FUS FUS +Arsenite (+As) HA (p62) HA (p62) 55

h i Input IgG HA(p62) IP Input IgG HA(p62) IP –Arsenite+Arsenite –Arsenite +Arsenite –Arsenite +Arsenite –Arsenite+Arsenite –Arsenite +Arsenite –Arsenite +Arsenite SMN SMN SMN SMN SMN SMN siRNA: Ctl Ctl Ctl Ctl Ctl Ctl 90 Ctl Ctl Ctl Ctl SMN siRNA: Ctl SMN Ctl SMN SMN SMN SMN FUS 34 SMN 63 HA (p62) HA (p62) SMN 55 34

Ctl siRNA j Pulldown k l m Peptide Peptide Peptide 100 SMN siRNA Input pulldown Input pulldown pulldown 0.001 80 ***

SMN input GST-p62 GST 60 SMN Ctl pept Sym Ctl Pept Sym p62 Ctl pept Sym 100 SMN p62 Input No pept Sym 75 GST-p62 63 p62 63 p62 40 35 SMN 63 p62 35 SMN SMN 20 GST 35 35 SMN 25 % cells with stress granules p62: + ++ 0 + + 30’ arsenite 30’ arsenite SMN: + + + +1 h recovery

Fig. 4 SMN mediates the interaction of p62 with FUS methylated by PRMT5 for degradation by autophagy. a Western blot of immunoprecipitates of endogenous proteins containing symmetrically dimethylated arginines from cells treated with arsenite (+As) or not (−As). b Western blot for proteins containing symmetrically dimethylated arginines in immunoprecipitates of FUS from cells stably expressing PRMT5 shRNA or control shRNA; top arrow indicates mass of FUS on western blots and bottom arrows indicate bands due to detection of immunoprecipitating IgG heavy and light chain. c Western blot of immunoprecipitates of FUS from cells with and without arsenite treatment (+As, −As); A non-speciﬁc band appears above PRMT5 in PRMT5 input western blots that is not depleted by shRNA targeting PRMT5 (see Supplementary Fig. 3e). d Left, autoradiograph of (3)H after in vitro methylation assay of GST or GST-FUS incubated with PRMT5-MEP50; Right, western blot of PRMT5 and GST in samples used for in vitro methylation assays (Right). e Western blot of HA-p62 immunoprecipitates in cells treated with DMSO (Ctl) or MTA (methylthioadenosine 1 mM, 48 h). f Western blot of FUS in HA- p62 immunoprecipitates from cells stably expressing PRMT5 or control shRNA. g Immunoﬂuorescent microscopy of GFP-SMN and TIAR (stress granule marker) in cells treated with arsenite. h Western blot of SMN in immunoprecipitates of HA-p62 from cells treated with SMN or control siRNA and either treated or not with arsenite. i Western blot of FUS in immunoprecipitates of HA-p62 from cells treated with siRNA targeting SMN or control (Ctl) and either treated or not with arsenite. j Western blot of recombinant SMN and GST-p62 from in vitro GST-pulldown assay. k Western blot of recombinant p62 and SMN from in vitro pulldown of control (Ctl) or symmetrically dimethylated peptides (Sym) incubated with recombinant p62 and SMN. l Western blot of recombinant p62 and SMN from in vitro pulldown of peptides incubated with recombinant p62 alone or p62 and SMN. m Percentage of cells transfected with control (Ctl) or SMN siRNA containing stress granules following 30 min arsenite treatment or following 1 h recovery from arsenite (n = 6, mean ± SD, Student’s t-test). HeLa cells were used in all experiments. Scale bar = 10 µm

2a–d). ALS patients with C9ORF72 repeat expansions have of antibodies recognizing symmetrically dimethylated arginines, reduced levels of C9ORF7217 and may therefore exhibit sig- staining was reduced in PRMT5-depleted cells (Supplementary natures of impaired p62-dependent clearance of proteins with Fig. 3h), and exhibited prominent staining of the nucleus, without symmetrically methylated arginines. evident cytoplasmic foci in non-ALS patients as expected (Fig. 6a, In patients with C9ORF72 expansions, inclusions were b). In both cerebellum and hippocampus, but not spinal cord of observed in the lumbar spinal cord containing TDP-43 and p62 patients with C9ORF72 repeat expansions inclusions containing as expected (Supplementary Fig. 6a, b). Validating the speciﬁcity p62 frequently also contained proteins with symmetrically

NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 dimethylated arginines (Fig. 6c–f, Supplementary Fig. 6b). While Together, this suggests that C9ORF72 and p62 form a complex we cannot exclude that p62 promotes symmetric dimethylation of that recognizes proteins like FUS containing symmetrically arginines of inclusions, with evidence above, this suggests that dimethylated arginines to eliminate them individually or as p62 is recruited to inclusions containing symmetrically dimethy- clusters in stress granules by autophagy (Fig. 7). Interruption of lated arginines in patients with C9ORF72 repeat expansions, but this process can cause accumulation of symmetrically methylated is unable to efﬁciently eliminate them, potentially due of the loss proteins as demonstrated in the mouse brain and in the of C9ORF72 (Fig. 2c, d). cytoplasm of patients with C9ORF72 repeat expansions.

HEK293T ab+Arsenite, +Bafilomycin A1 IgG IP HA(FUS) IP InputIgG IP HA(FUS) IP

FUS: WT R216C R218K R216C+218KR521G R521H 100 75 FUS 63

FUS WT FUS R218K FUS FUS WT FUS R218K FUS

FUS R218K FUS FUS WT FUS R218K FUS WT FUS WT FUS R218K FUS Tubulin 48 75 HA (FUS) SYM 63 p62 IgG Heavy ** chain SMN 18 ** 75 * FUS 16 ** Nucleus 14 Cytoplasm 12 c 10 8 6 4

(+Arsenite, + Bafilomycin A1) 2 p62-FUS PLA signal counts 0 FUS:WT R218K R216C R216C R521G R521H + R218K p62-FUS WT p62-FUS R218K p62-FUS R521G d 0.003 ** 0.7 0.0025 0.6 ** 0.5

+Arsentite +Bafilomycin A1 0.4 0.3

Proximity ligation assays (p62-FUS) 0.2 0.1 0 p62-FUS R216C p62-FUS R216C+R218K p62-FUS R521H FUS: WT R216C R218K R216C R521G R521H

Fluorescence intensity, (+As, +Baf A1) + R218K Cytoplasmic / nuclear ratio of FUS

e 0.00015 R216C R216C *** FUS: FUS: 2.0 WT R216C R218K +R218K WT R216C R218K +R218K DMSO BAF DMSO BAF DMSO BAF DMSO BAF DMSO BAF DMSO BAF DMSO BAF DMSO BAF DMSO BAF 1.5 100 Tubulin 75 48 1.0 FUS 63 17 LC3-I

Relative FUS levels 0.5 Tubulin 48 LC3-II 0 WT R216C R218K R216C+R218K

Primary mouse motor neurons fgNo arsenite Cerebellum Hippocampus –/– 1h arsenite WT p62 p62: WT KOWT KOWT KO WT KO WT KO 180 1h arsenite + 30’ recovery 135 100 60 0.003 75 NS ** 63 48 SYM 1h As

35 40 TIAR TIAR 25 20

20 75 FUS

63 Tubulin

% cells with stress granules 48 0 TIAR TIAR –/– 63 p62

WT p62 1 h As, 30 min recovery

* p = 0.09 40 NS h 50 i 40 30 jk60 *** * * *** *** 40 40 30 30 20 40 30 30 20 20 20 20 10 10 10 20 10 10 inclusion (%) inclusion (%) inclusion (%) inclusion (%) inclusion (%) inclusion (%) 0 0 0 0 Zchc6 Exon 27 Zcchc6 Exon 27 0 0 Zcchc6 Exon 27 Zcchc6 Exon 27 Zcchc6 Exon 27 Zcchc6 Exon 27 Ctl p62 wt p62 KO wt p62 KO wt p62 KO Ctl Fus Ctl p62 + FUS HA-FUS siRNA siRNA siRNA siRNA siRNA siRNA Spinal Cord Cerebellum Striatum MN1 MN1 MN1

1.0 –/– lmnAll detected mRNAs 62 ** p 0.8 Putative p62 –/– ** mRNAs mis-spliced in p62 * splicing Putative FUS 1.2 ** 0.6 *** targets splicing 1.1 targets 0.4 1 97 862 mRNAs containing high levels of FUS-motifs inclusion (%) 0.9 1386 p = 0.03 Zcchc6 Exon 27 Lagier-Tourenne, C. Fraction of mRNAs 0.2 WT R216C R218K R216C R521G R521H Kurosawa M. et al., 2015 et al., 2012 Fus siRNA + FUS FUS FUS R218K FUS FUS putatively mis-spliced in FUS p = 1.57 × 10–7 0.0

–/– – log10(p value of putative alternative splicing in p62 )

10 NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 ARTICLE

Discussion achieve these effects by repeatedly eliminating parts of stress Canonically, autophagy receptors bind ubiquitin on cargoes to granules, rather than engulfing and degrading entire stress direct their degradation by selective autophagy42, but this is not granules in a single autophagosome. Indeed, p62 bound to SMN always the case4. While ubiquitin was not enriched in stress and FUS both in the presence and absence of oxidative stress granules in our analyses, we cannot exclude that elimination of when stress granules are rare (Figs. 1n, 4h, i). Additionally, while stress granules requires ubiquitination in addition to symmetric methylation of FUS alone is insufficient to direct degradation of arginine methylation. Recent reports highlight the ability of stress granules (Supplementary Fig. 5f), it is sufficient to cause alternative post-translational modifications to target cargoes for degradation of FUS by autophagy (Fig. 5e). Together, this sug- autophagic degradation. For example, P granules, an RNA- gests that turnover of submicroscopic complexes of FUS and containing structure unique to C.elegans germline cells, are possibly individual FUS proteins may constitutively occur by p62- eliminated by autophagy43. Notably, this process is distinct from dependent autophagy and the clustering of multiple SYM into that described here in that it requires asymmetric, not symmetric large granules during stress renders this process more readily arginine methylation of P granule proteins, and is independent of observable with a microscope. the C. elegans p62 homolog8. Arginine-methylated proteins are Multiple RNA surveillance mechanisms patrol the cytoplasm enriched in RNA-binding domains36,44. In the absence of an to prevent translation of mRNAs that contain errors or are enzyme known to remove arginine methylation from proteins in unspliced49,50. Symmetrically methylated proteins are highly cells, autophagy is one mechanism that enables turnover of enriched in the nucleus and many are involved in RNA proces- arginine methylated proteins regrouped in diverse RNA granules. sing. Our data suggests that symmetrically methylated proteins PRMT1 asymmetrically methylates arginines on FUS31 causing that escape into the cytoplasm are targeted for degradation by FUS mutants associated with ALS to accumulate in the cytoplasm p62-dependent autophagy. The mechanism we describe may be a in stress granules45. Similarly, PRMT1-mediated methylation of constitutive RNA surveillance and quality control mechanism FMRP and other proteins drives their recruitment to stress that detects escape of nuclear RNA processing complexes into the granules46. The current data suggest that an independent type of cytoplasm and restricts production of toxic proteins from arginine methylation by PRMT5 at FUS R218 controls a sub- immature mRNA or mRNA with errors. sequent step, the targeting of cytoplasmic FUS for degradation by One model suggests that stress granules or a related derivative autophagy. underlie the pathology of at least some forms of ALS13. Mutations ALS-linked C9ORF72 with repeat expansions that decrease in another protein genetically linked to ALS, VCP, inhibit its levels of endogenous C9ORF72 were previously shown to increase normal function in promoting elimination of stress granules by stress granule number, but the mechanism remained unknown47. autophagy7. Our data demonstrates a cascade of events governing Mimicking these conditions with a partial knockdown of the recognition and degradation of stress granules that requires a C9ORF72, we observed a defect in elimination of stress granules protein genetically linked to ALS at each step (FUS, SMN, p62, (Fig. 2c, d). C9ORF72 also has a role in controlling the initiation C9ORF72) (Fig. 7). This reinforces a unifying model in which of phagophore formation by interacting with Rab1a48. Intrigu- mutations that instigate cytoplasmic accumulation of RNA spli- ingly, Rab1a was identified in stress granules22. We found that cing proteins or impinge on their elimination by autophagy can C9ORF72 and p62 decorated stress granules to initiate their culminate in ALS. elimination by autophagy. This suggests C9ORF72 and p62 are Interestingly, multiple proteins modified with dimethyl- part of a broader complex, potentially including Rab1a at stress arginines associated with both C9ORF72 and p62. Proteins granules to promote engulfment of stress granules or fragments bearing these dimethyl sites included several proteins genetically thereof within autophagosomes. At the same time, we cannot linked to ALS, such as FUS, HNRNPA1, and TAF15 (Supple- exclude that C9ORF72 effects independent of p62, or p62 func- mentary Table 6). This suggests that arginine dimethylation of tions independent of autophagy also promote elimination of these multiple ALS-linked substrates may drive association with stress granules. p62 and elimination of stress granules. As we describe for ALS- Electron microscopy exhibited p62 and FUS co-localized on linked FUS mutants (Fig. 5a–c, e), it is possible that some ALS- small electron-dense clusters (~20–80 nm) on the periphery of linked mutations in these other proteins disrupt the ability of p62 stress granules (Fig. 2h–k) and p62 is required to eliminate stress to recognize and eliminate stress granule-linked substrates granules (Fig. 2a, b, h–m). It is possible that p62 and autophagy by autophagy.

Fig. 5 p62 regulates turnover of FUS and splicing of its targets. a Western blot of symmetrically dimethylated arginines in FUS immunoprecipitates from HEK293T cells. b Left, western blot of FUS immunoprecipitates from HeLa cells treated with arsenite and Bafilomycin A1; Right, western blot of FUS in HeLa cells in conditions used for PLA in (c). c Left, PLA of p62 and wild type or indicated FUS mutants in HeLa cells treated with arsenite and Bafilomycin A1; Right, quantification of PLA; images counted (>10 cells/image): wild type (11), R218K (9), R216C (8), R216C + R217K (8), R521G (9), R521H (9) (n = 3, mean ± SD, Student’s t-test). d Quantification of nuclear/cytoplasmic distribution of wild type and mutant FUS by fluorescence intensity in HeLa cells treated with arsenite and Bafilomycin A1; cells counted: wild type (32), R218K (19), R216C (8), R216C + R218K (16), R521G (31), R521H (21) (mean ± SD, Student’s t-test). e Left, western blot of FUS wild type, indicated mutants and LC3 upon Bafilomycin A1 treatment of HeLa cells; Right, quantification of western blots normalized to Tubulin (n = 6, mean ± SD, Student’s t-test). f Left, percentage of cells containing SGs in primary motor neuron cultures of wild type and p62−/− mice (n = 2–4); 35–150 cells were counted per condition (mean ± SD, Student’s t-test); Right, representative images. g Western blot of lysates from wild type (WT) and p62−/− (KO) mice. h–l RT-PCR analysis of Zcchc6 splice variants in (h) p62 depleted MN1 cells (n = 4), i indicated tissues of WT and p62−/− mice (n = 3), j FUS depleted and overexpressing MN1 cells (n = 4), k FUS and p62 depleted MN1 cells (n = 3), and l MN1 cells depleted of endogenous FUS with siRNA targeting FUS 3’UTR and expressing FUS wild type or indicated mutants (n = 4, mean ± SEM, Student’s t-test). m Venn-diagram depicting overlap between mRNAs putatively alternatively spliced in p62-/- brains and those bound to FUS and mis-spliced in FUS-/- −/− brains. n A plot of the fraction of mRNAs putatively mis-spliced in p62 brain (y-axis) vs. the increasing significance of this (-log10 p-value, x-axis); black line—all detected mRNAs, green line—mRNAs enriched in FUS-binding motifs, red line—only mRNAs putatively mis-spliced in p62−/− brain. p = 0.0357 that putative mRNAs mis-spliced in p62−/− brains are enriched in FUS-binding motifs. Scale bar = 10 µm

NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7

a b Cerebellum

Non-ALS IgG Ctl SYM p62 Non-ALS Merge cd Cerebellum

C9-ALS IgG Ctl SYM p62 C9-ALS Merge e SYM p62 C9-ALS Merge f 80

40 Hippocampus 20 SYM p62 C9-ALS Merge

% p62+ inclusions enriched in 0 symmetric arginine dimethylation SCSC Hc Cb sALS C9-ALS

Fig. 6 Patients with C9ROF72 repeat expansions exhibit inclusions rich in symmetrically arginine-dimethylated proteins. a, b Immunofluorescent microscopy of non-ALS human cerebellum labeled with (a) control non-specific antibodies or (b) antibodies specific for p62 and symmetrically dimethylated arginines. c, d Immunofluorescent microscopy of cerebellum of patient with C9ORF72 repeat expansions labeled with (c) control non-specific antibodies or (d) antibodies specific for p62 and symmetrically dimethylated arginines; arrows highlight inclusions containing p62 and symmetrically dimethylated arginines. e Immunofluorescent microscopy of hippocampus of patient with C9ORF72 repeat expansions labeled with antibodies specific for p62 and symmetrically dimethylated arginines; arrows highlight inclusions containing p62 and symmetrically dimethylated arginines. f Quantification of fraction of p62 inclusions positive for symmetrically dimethylated arginine proteins as observed in spinal cords (SC), hippocampi (Hc) and cerebellums (Cb) of patients with C9ORF72 repeat expansions (C9-ALS) and sporadic ALS (sALS). Scale bar = 10 µm. sALS Spinal Cord (n = 2); C9 ALS Spinal Cord (n = 2); C9 Hippocampus (n = 1); C9 Cerebellum (n = 3), mean ± SD, Student’s t-test. Cytoplasmic p62 inclusions were counted across 3–10 frames (10–80 p62 inclusions) in each sample

Repeat expansions in C9ORF72 are the most common genetic Care Committee. Information of p62 WT and p62−/− mice used in the study is cause of ALS15,16. Overexpression of dipeptide repeats translated included in Supplementary Table 8. from these repeat expansions causes cytoplasmic accumulation of RNA-binding proteins usually concentrated in the nucleus and Isolation of mixed motor neurons. Mice embryos were collected from pregnant can initiate the formation of inclusions51,52. However, repeat mice between day E13.5 and E14.5. Spinal cords and cortices were dissected using Zeiss Stereo Discovery V20 microscope (Carl Zeiss, Oberkochen, Germany). Clean overexpression does not replicate all facets of the spinal cord spinal cords were then placed in dissection buffer (sucrose 40 g/L, dextrose 1 g/L, motor neuron pathology and other symptoms observed in ALS and HEPES 2.4 g/L in PBS), minced with scissors and incubated with trypsin patients53. Cells depleted of C9ORF72 to mimic reduced (Sigma-Aldrich) for 30 min at 37 °C. The cells were separated using a 1 mL pipette C9ORF72 expression in patients17 were incapable of targeting and placed in S.C. NFeed (MEM/HBSS, Hyclone, hormones (insulin 10 μg/mL, transferrin 200 μg/mL, BSA 10 μg/mL, putrescine 32 μg/mL, selenium 26 ng/mL, stress granules for p62-dependent autophagy (Figs. 1, 2). In ALS T3 20 ng/mL, hydrocortisone 9.1 ng/mL, progesterone 13 ng/mL, and NGF patients with C9ORF72 repeat expansions symmetrically 5 ng/mL) Sigma-Aldrich, Horse serum 1.3%, Multicell and ABAM 1%, Multicell) arginine-methylated proteins accumulate in cytoplasmic foci neuronal culture media. The cells were then counted and seeded at 300,000 cells/ often co-labeled with p62 (Fig. 6d–f). well in 12-well plates pre-treated with poly-D-Lysine 1 mg/mL (Sigma-Aldrich) and Matrigel 0.5% (S.C., Corning, VWR). After four to seven days, mitotic cells Our evidence, thus suggests that loss of C9ORF72 may be were killed using Arabinofuranosylcytosine 1.4 μg/mL (S.C., Calbiochem). required to generate the complete spectrum of ALS pathology, by preventing stress granule components from being eliminated and Cell culture, transfections, and treatments. HeLa (CCL2, ATCC) and MDA- thereby contributing to inclusion formation, misregulation of MB-231 (HTB-26, ATCC) were cultured in Dulbecco’s Modiﬁed Eagle’s Medium splicing and attendant pathology in these ALS patients. containing 10% fetal bovine serum, 2 mM L-glutamine and penicillin- streptomycin. MDA-MB-231 (HTB-26, ATCC) cells were genetically deleted of ATG5 using CompoZr Zinc Fingers technology (Sigma-Aldrich). Motor-neuron- derived MN-1 cells were cultured in DMEM (GIBCO) supplemented with 10% Materials and methods μ − − fetal bovine serum. All stable lines were maintained in 2 g/ml of puromycin. Animal studies. p62 WT and p62 / mice were a gift from Dr. Herbert W. Virgin For transient transfection with DNA, Lipofectamine2000 (11668-019, (Washington University—School of Medicine)54. Animal husbandry was per- Invitrogen) was used. Silencer Select siRNAs (Life Technologies) were transfected formed in accordance with federal guidelines and the University of Ottawa Animal at 10 nM concentration with RNAiMax (13778150, Invitrogen). The cells were

12 NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 ARTICLE

PRMT5

C9ORF72 SDMA Phagophore SMN p62 LC3 FUS

m7Gppp SDMA

m 7 Gppp AAA m7Gppp AAA AAA m 7 Gppp m

7 m 7 Gppp SDMA 7 Gppp

m Gppp AAA

SDMA

AAA

Gppp

7 Gppp AAA SDMA AAA SDMA SDMA

AAA

Gppp 7 AAA m SDMA AAA

SDMA m7Gppp AAA Clearance by selective autophagy

SDMA

AAA AAA

Gppp 7 AAA

m AAA 7 SDMA Gppp AAA m AAA

Stress granule SDMA m Gppp

7 Gppp 7 Gppp 7 m m 7 m Gppp

7 Gppp SDMA m SDMA

m 7 Gppp Gppp AAA 7 AAA m AAA

SDMA

m AAA 7 Gppp AAA

m AAA SDMA 7 Gppp

m 7 SDMA Gppp

SDMA 7 Gppp m SDMA

SDMA AAA

7 Gppp

Gppp 7 m

AAA AAA m

AAA

AAA SDMA AAA AAA

SDMA

m Gppp

7 Gppp 7 Gppp 7 m m 7 m Gppp

7 Gppp SDMA m SDMA

m 7 Gppp Gppp AAA 7 AAA m AAA

SDMA

m AAA 7 Gppp AAA

m AAA SDMA 7 Gppp

m 7 SDMA Gppp

SDMA 7 Gppp m AAA SDMA

SDMA AAA AAA SDMA AAA

7 Gppp

Gppp AAA 7 m

AAA

AAA m SDMA m Gppp

7 Gppp 7 Gppp 7 m m 7 m Gppp

7 Gppp SDMA m SDMA

m 7 Gppp Gppp AAA 7 AAA m AAA

SDMA

m AAA 7 Gppp AAA

m AAA SDMA 7 Gppp

m 7 SDMA Gppp

SDMA 7 Gppp m SDMA

SDMA AAA

7 Gppp

Gppp 7 m

AAA AAA m

Autolysosome

Fig. 7 Model. Stress granule proteins such as FUS are symmetrically dimethylated on arginines by PRMT5 to bind Tudor-domain containing protein SMN for subsequent recruitment and degradation by p62-C9ORF72 mediated selective autophagy harvested for analyses 48 h post transfection. siRNAs and plasmids used in the min. At least 100 cells were counted per experiment for quantiﬁcation of cells study are listed in Supplementary Tables 9, 10. containing stress granules. HeLa and MN-1 cells were treated with Baﬁlomycin For all stress granule experiments, HeLa and MDA-MB-231 cells were A1 (196000, EMN Millipore) for 16 h at 400 nM. HeLa cells were treated with (i) induced with 0.5 mM sodium arsenite (71287, Sigma-Aldrich) for 30 min Chloroquine (sc-205629A, Santa Cruz Biotechnology) for 16 h at 20 μM, (ii) followed by a PBS wash and recovery in complete DMEM up to 1 h. Primary MTA (Sigma-Aldrich) for 48 h at 1 mM, (iii) PRMT5 inhibitor—EPZ015666 cultures of mixed motor neurons were induced with 0.5 mM sodium arsenite (SML1421, Sigma-Aldrich) for 48 h at 3 μM, and (iv) Ubiquitin E1 inhibitor— (71287, Sigma-Aldrich) for 1 h followed by a PBS wash and recovery up to 30 PYR-41 (N2915, Sigma-Aldrich) for 3 h at 50 μM.

NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7

fl Lentiviral vector production and transduction. HeLa cells stably expressing Aldrich). For immuno uorescence, the cells were subsequently washed thrice in shRNA targeting PRMT5 (sh(PRMT5)) or luciferase (sh(luciferase)) were gener- PBS and incubated with 1/300 dilution of secondary antibodies (goat anti-mouse ated via lentiviral transduction. Lentivirus was produced through transfection of Alexa Fluor 488 or 546, anti-rabbit Alexa Fluor 488 or 546 or donkey anti-goat 293T cells with the packaging plasmid pLKO.1-sh. The media was harvested 48 h Alexa Fluor 488 (Invitrogen)) for 1 h. After three PBS washes, cells were mounted post transfection and filtered through a 0.45 μM PES membrane. HeLa cells were with Vectashield media (H-1200, Vector Laboratories) and imaged on a Zeiss fl seeded in 10 cm plates (250,000 cells) 24 h before infection. The cells were infected AxioObserver Z1 epi uorescent microscope, a Zeiss 510 confocal microscope or a with the lentivirus containing media along with 8 μg/mL of polybrene. After 48 h, Zeiss LSM880 AxioObserver Z1 confocal microscope. At least three images were media was changed and cells were selected with puromycin (2 μg/mL). Equivalent taken for each experiment. amounts of virions encoding shRNA targeting luciferase as a control was used to transduce HeLa cells. Knockdown in stable lines was validated by RT-PCR and Quantification of cytoplasmic inclusions in ALS patients. P62 inclusions in the immunoblotting. cytoplasm were manually counted across 3–10 frames of images per n. About 10–80 P62 inclusions were counted per n. The fraction of these P62 inclusions DNA constructions. The plasmid pDEST-FUS-WT was obtained from Addgene positive for symmetric arginine dimethylation was then manually ascertained. (#26374). Three different arginine (R) mutations were generated: (i) arginine to cysteine at position 216 (R216C), (ii) arginine to lysine at position 218 (R218K), Quantification of microscopic data. Immunofluorescent images of punctae or and (iii) both mutations (R216C and R218K) by PCR using oligos indicated in the stress granules containing P62, C9ORF72, and TIAR were first segmented using primers table. PCR reactions were digested with DPNI for 1 h and transformed in 57 α fi SQUASSH . Segmented objects were then binarized prior to counting them using DH5 . DNA was puri ed from colonies and sent for sequencing. Constructs that ‘Analyze Particles’ function of Fiji. Binarized images of objects from one channel were successfully mutated were then purified by MaxiPrep (Qiagen) as per man- ’ were intersected with respective binarized objects from the other channel using ufacturer s guidelines. ‘AND’ function of Fiji. The resulting number of overlapping punctae were then used to calculate % overlap. PLA dots were quantified using Duolink® ImageTool Reverse transcription (RT)-PCR and quantitative RT-PCR. After RNA extrac- (DUO90806, Sigma-Aldrich). tion with Trizol, RT-qPCR was performed with Superscript II Reverse Tran- scriptase (Invitrogen) and GoTaq qPCR Master Mix (Promega, A6002). Fold 3D reconstruction of Z-stacks μ ΔΔ . Z-stack images separated by 0.2 m were acquired change of transcripts was calculated using the Ct method, normalized to on a Zeiss LSM880 AxioObserver Z1 confocal microscope with Airyscan. 3D 18SrRNA. Sequences of primer pairs used in the study are listed in Supplementary surfaces were rendered from Z-stack image files using Imaris. Table 11. For RT-PCR, cDNA was synthesized with random hexamers and Promega AMV cDNA synthesis (Promega, M5101). PCRs were then set up using Promega Electron microscopy. For transmission electron microscopy, cells were fixed in GoTaq DNA Polymerase (M5101) according to manufacturer’s instructions. RT- 2.5% glutaraldehyde in a 0.1 M sodium cacodylate buffer and then incubated in PCR conditions were as follows for validation of splicing targets: 95 °C for 2 min, OsO4 (2%) in a 0.1 M sodium cacodylate buffer. After washing in 0.1 M sodium (95 °C for 30 s, 55 °C for 30 s, and 72 °C for 45 s) × 32 cycles, and 72 °C for 10 min. cacodylate the cells were dehydrated in increasing concentrations of alcohol. The Specific RT-PCR conditions used for particular primers are available upon request. most concentrated alcohol was replaced with acetone. The material was permea- Amplicons were run on a 2% agarose gel (containing 5 μL of EtBr 20 mg/mL) and bilized with increasing concentrations of Araldite diluted in acetone. Finally, the visualized under UV light. For analysis of splice variants, densitometry was tissues were embedded in Araldite (Huntsman Advanced Materials LLC, United performed to measure percentage exon inclusion with or without normalization to States). Ultrathin sections (80 nm) were prepared using an Ultracut Leica UC6 wild-type condition. ultramicrotome (Leica Microsystems, Germany) and placed onto copper grids coated with Formvar film. Sections were stained with uranyl acetate and lead citrate solutions and examined with a transmission electron microscope (JEOL JEM 1230, Isolation of autophagosomal fractions. Autophagosomes were fractionated as 30 fl 2 Japan). described . Brie y, 20 160 cm plates of HEK293 cells were harvested in 0.25 M For immungold labeling, the cells were fixed in 4% paraformaldehyde and 0.5% sucrose 10 mM HEPES pH 7.2, washed several times in the same solution, trans- glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4). After dehydrating the ferred into 0.25 M sucrose containing protease inhibitor cocktail (Roche, EDTA- samples with increasing concentrations of ethanol, the pellets were embedded in free), and disrupted by nitrogen cavitation (Parr Instruments, 100 p.s.i. 30 s, 50 p.s. LR White Resin. The blocks were sectioned with ultracut (Leica EM UC 6) using a i. 30 s, and 25 p.s.i. 8 min). Lysate was centrifuged twice at 2000× g to eliminate diamond knife. The thin sections were mounted on Formvar-coated nickel grids. intact cells and large debris. Supernatant was then centrifuged at 17,000× g (12 The grids were floated on drops of 1% BSA/0.01% Tween 20 for 1 h, then incubated min) to pellet organelles. The pellet was resuspended in 0.25 M sucrose 10 mM for 1 h with primary antibody or a mixture of rabbit and mouse primary antibody HEPES, pH 7.2, and treated with a mixture of RNAse A/T1 (Fermentas) for 10 diluted with PBS-0.05% Tween 20 (PBST) diluted 1/200. Next, the samples were min, 37 °C. Organelles were then loaded at the bottom of a discontinuous Histo- washed in PBST and incubated for 2 h with a mixture of secondary antibody in denz density gradient (26, 24, 20, and 15%) and centrifuged at 90,000× g for 3 h in – – PBST. The secondary antibodies were 12-nm colloidal gold-conjugated goat anti- an SW41 rotor (Beckman). Fractions at the interface of 15 20% and 20 24% rabbit IgG antibody (111-205-144, Jackson) and 18-nm colloidal gold-conjugated densities are enriched in autophagosomes and autophagolysosomes, respectively. goat anti-mouse IgG antibody (115-215-068, Jackson) that were applied at a Autophagosome-enriched fractions were used for all analyses. dilution of 1/50 for 2 h at room temperature. For control staining the primary antibody was omitted and secondary only was used. After antibody labeling, the Immunofluorescence and proximity ligation assay. Five micron sections were grids were washed three times in PBST, rinsed in distilled water, stained with cut from formalin-fixed, paraffin-embedded blocks of cerebellum and hippo- uranyl acetate and lead citrate, and imaged with a transmission electron campus and mounted onto coated glass slides. Paraffin-embedded sections were microscope (JEOL JEM 1230, Japan). deparaffinized in xylene (3 × 5 min) and rehydrated through graded EtOH to deionized water: 100% EtOH (2 × 5 min), 95% EtOH (2 × 5 min), 80, 70, and 50% Quantification of electron microscopy. Forty-six electron microscopy images EtOH (1 × 5 min). Additionally, frozen tissues were sectioned and mounted onto containing any signal for P62 or FUS were collected at random from the cytoplasm coated glass slides prior to fixation in 4% PFA in PBS (10 min). Antigen retrieval of cells and 193 electron-dense clusters, autophagolysosomes or cytoplasmic labels was performed in 0.05% Tween 20, 10 mM sodium citrate buffer pH 6.0 (prepared were counted. P62 and FUS staining was evaluated as being associated with with 0.1 M citric acid) using a decloaking chamber (Biocare Medical, Walnut electron-dense stress granule-like structures whenever it or they were directly Creek, CA,USA) at 125 °C for 30 s and 90 °C for 10 s. The sections were cooled to localized within a cluster of electron-dense material of at least 100 nm resembling room temperature and rinsed with deionized water five times and PBS once. They structures labeled in other cells with stress granule markers FUS and FMRP. P62 were permeabilized in 0.2% Triton-X-100 in PBS for 45 min, washed three times in and FUS was considered as within autophagolysosomes if it was within a trans- PBS, blocked in 10% normal goat serum (S-1000, Vector Laboratories) in PBS for 1 lucent membrane-bound structure of >200 nm containing heterogenous contents. h, and incubated with primary antibodies in blocking solution overnight at 4 °C. If more than one gold bead for P62 or FUS was associated with a single structure it The sections were then washed thrice in PBS, incubated in fluorescently labeled was counted as a single co-localization event. P62 and FUS not localized to secondary antibodies (1/250 dilution in PBS for 1 h), washed thrice in PBS, electron-dense clusters or membrane-bound organelles was labeled as cytoplasmic. incubated in 0.5 μg/ml DAPI (D9542, Sigma-Aldrich) in PBS for 10 min, washed If P62 and FUS were within 100 nm of each other in the cytoplasm they were twice in PBS, and mounted with ProLong Diamond Antifade (P36961, Invitrogen). considered co-localized. Cells were prepared for microscopy as previously described55,56. The cells were grown on 0.17 mm glass coverslips prior to fixation in 4% PFA in PBS (10 min). The cells were rinsed in PBS and incubated 10 min in 0.2% Triton-X-100 in PBS Co-immunoprecipitation. Cells were washed in cold PBS and lysed in NP40 lysis containing 20 mM NH4Cl. After washing with PBS, the cells were blocked in 5% buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% NP40, Roche Complete Pro- milk in PBS for 1 h, washed in PBS, and incubated with 1/200 dilution of primary tease Inhibitor Cocktail Tablet EDTA free) for 30 min at 4 °C and centrifuged at antibody or control non-specific antibody overnight at 4 °C. For PLA, the cells were 1000× g, 5 min prior to immunoprecipitation. Mouse brain and spinal cord tissues subsequently processed as per manufacturer’s instructions (DUO82049, Sigma- were homogenized in NP40 lysis buffer using Dounce homogenizer and

14 NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 ARTICLE centrifuged at 6000× g, 10 min at 4 °C. For IP, 500 μg lysate was pre-cleared with Abnova) for 1 h at 4 °C. Upon precipitation, the mixtures were washed five times 5 μl (0.15 mg) protein G-Dynabeads (10004D, Life Technologies) for 20 min at 4 °C with buffer A, 2 min spins at 900× g. and estimated for protein content (DC Protein Assay, Bio-Rad). Every 500 μg lysate Recombinant GST or GST-p62 was bound to glutathione-agarose beads in was incubated with 1 μg antibody for 3 h at 4 °C before incubation with 20 μl wet buffer B (50 mM Tris pH 7.5, 200 mM NaCl, 0.1% triton-X-100) by end-to-end volume of protein G-Dynabeads (1.5 h, 4 °C). Alternatively, lysates with GFP or toppling at 4 °C for 1 h 30 min. Beads were previously blocked with 3% BSA in PBS Myc-tagged proteins were directly incubated with 20 μl GFP-Trap®_A slurry (gta- (overnight at 4 °C). Following the incubation, bound beads were spun down at 20, ChromoTek) or 50 μl wet volume of anti-c-myc magnetic beads (PI88842, 9000× g, 5 min, 4 °C, and washed three times for 5 min each with buffer B, 1 min Thermo Fisher Scientific) for 2 h at 4 °C. Beads were collected by centrifugation or spins at 9000× g. Recombinant SMN was pre-cleared with blocked glutathione- using a magnetic support (12321D, Thermo Fisher Scientific) and washed with 200 agarose beads bound to GST for 20 min at 4 °C. Following centrifugation at 9000× μl lysis buffer three times. SDS-sample buffer was added and the samples were g, 5 min, 4 °C, pre-cleared supernatant was collected and incubated with previously boiled at 99 °C for 5 min prior to western blot analysis. prepared glutathione-agarose beads bound to GST or GST-p62 for 1 h at 4 °C. For RNAse treatments, equal amounts of lysate were untreated or treated with Upon precipitation, the mixtures were washed five times with buffer B, 1 min spins RNase A/T1 Mix (FEREN0551, Fermentas, 28 µg/ml RNase A and 70 U/ml RNase at 9000× g. T1) for 15 min at 37 °C prior to antibody incubation. RNA digestion was confirmed The bound proteins were then analyzed by SDS-PAGE. on a 1% agarose gel by electrophoresis.

— Quantification of co-immunoprecipitation. Densitometry-based analyses were Peptide-pulldown assay. The peptide KGRGRGRGRG was custom ordered employed. For each protein in an experiment, an area encompassing the maximum from GenScript in two forms: all 4 arginines (i) symmetrically dimethylated, or (ii) μ band size was selected and mean intensity of this area was measured using the asymmetrically dimethylated. A volume of 1 mg of each of the peptides in 400 lof μ fi ‘Histogram’ function in Photoshop. Mean intensities of interacting proteins from PBS was incubated with 400 lofAf -Gel 10 (1536099, Bio-Rad) media washed in control IgG immunoprecipitates were subtracted from those of respective test cold deionized water to make 50% slurry. The uniform suspension was left to rotate μ immunoprecipitates followed by normalization to the mean intensity of the target on a wheel overnight at 4 °C. A volume of 40 l of 1 M ethanolamine HCl in 0.5 M of the immunoprecipitating antibody. Tris pH 8 was added to the slurry to block any active esters and left rotating for 2+ h at 4 °C. Peptide bound media was then washed in 500 μl PBS thrice, resuspended in 1 ml of 0.1% sodium azide in PBS, and stored at 4 °C. For the pulldown Whole cell extracts and western blotting. Cells were washed in cold PBS and assays, affi-gel 10 media—10 μl per reaction was washed in PBS thrice and lysed in 5 mM Tris pH 7.4, 75 mM NaCl, 0.5 mM EDTA, 0.5% Triton-X-100 with resuspended in 1 ml of binding buffer (20 mM HEPES pH 7.9, 200 mM NaCl, 10% protease inhibitors. Lysate was then spun down at 1000× g, 5 min at 4 °C to Glycerol, 0.05% Tween 20) prior to pre-clearing 375 ng of recombinant p62 and/or estimate the protein content of supernatants. Where noted, the cells were lysed in SMN for 20 min at 4 °C. Pre-cleared proteins were separated from the beads upon RIPA lysis buffer (10 mM Tris pH 7.4, 100 mM NaCl, 1 mM EDTA, 1% NP-40, centrifugation at 9000× g, 5 min. As a positive control, 1% of the supernatant was 0.5% NaDOC, 0.1% SDS, 10 µg/ml PMSF with protease inhibitors) and spun down frozen; remaining was incubated with 30 μlofaffi-gel 10 media bound to (i) no at 15,000× g, 10 min at 4 °C. Mouse cerebellar and hippocampal tissues in NP40 peptide, (ii) symmetrically dimethylated, or (iii) asymmetrically dimethylated RG lysis buffer were vortexed vigorously with two stainless steel beads per tissue (x4) peptide, on a rotating wheel for 1 h at 4 °C. Mixture was then washed six times, (SSB32, Next Advance) every 10 min for 1 h, followed by end-to-end toppling at each in 500 μl of binding buffer with 5 min rotations at 4 °C and 9000× g, 1 min 4 °C for 1 h. Homogenates in NP40 lysis buffer were then centrifuged at 6000× g, spins at 4 °C. SDS-sample buffer was added and the reactions were boiled at 99 °C 10 min at 4 °C. For western blotting, equal amounts of protein samples were for 5 min prior to western blot analysis. resolved on 10% (w/v) acrylamide gels, transferred to PDVF membrane (IPVH00010, EMD Millipore), blocked with 5% milk in TBST (10 mM Tris-HCl pH 8, 150 mM NaCl, 0.05% Tween 20) for 1 h and probed with primary antibody In vitro methylation assay in TBST overnight at 4 °C. All antibodies were used at 1/1000 dilution. Blots were . then washed in TBST, probed with HRP labeled secondary antibodies (goat anti- a. Substrate and enzyme preparation: GST-FUS-WT (44978, Addgene) and mouse, goat anti-rabbit and donkey anti-goat), washed again in TBST and imaged pGST2 with no insert were transformed in BL21 strain. Colonies were picked with HRP substrate (WBLUR0100A, EMD Millipore) on an ImageQuant LAS to grow overnight cultures in 50 ml LB at 37 °C. Culture volumes were 4000 system (GE Healthcare). Mean intensity levels were measured for increased to 1000 ml and grown at 37 °C until OD 0.5. Protein expression was densitometry-based normalization to tubulin levels. For each protein in an induced with 0.1 mM IPTG for 5 h at 37 °C. Cultures were spun down at experiment, an area encompassing the maximum band size was selected and mean 1700× g for 20 min, pellets lysed in PBS containing protease inhibitors, ‘ ’ intensity of this area was measured using the Photshop Histogram function. Full sonicated five times 15 s each with 10 s intervals at power 3, and centrifuged at size uncropped blots are shown in Supplementary Fig. 7. Antibodies used in the 15,309× g for 20 min at 4 °C. Supernatants were solubilized in 1% Triton-X- study are listed in Supplementary Table 12. 100 for 30 min, 4 °C and incubated with 500 μl GST beads (16100, Pierce) overnight at 4 °C. Washed beads were then stored in PBS at 4 °C. Active p62 and SMN protein purification. P62 and SMN1 expressing plasmids were human PRMT5 in complex with MEP50 was purchased (SRP0145, Sigma- ordered from GenScript with the codon optimized for overexpression in E. coli. Aldrich). The DNA constructs contain a BamH1 and a Xho1 restriction site flanking each b. Enzymatic reactions: Equal amounts of protein on agarose beads were each μ μ gene. After amplification of the vectors, the plasmids were digested with BamH1 incubated with 1 g PRMT5-MEP50 and SAM in a 30 l reaction buffer (50 and Xho1 and the DNA fragments were sub-cloned in a vector enabling the mM Tris pH 8, 1 mM DTT, 10 mM NaCl) for 1 h at 30 °C and 3 h at 37 °C. expression of P62 and SMN1 as TEV cleavable GST fusion proteins58. pGST2-p62 For western blot and mass spectrometry analyses, 80 uM cold SAM (A7007, and pGST2-SMN1 were transformed in Rosetta cell and plated on amp- Sigma-Aldrich) was used while 1 µCi SAM[3 H] (NET155V250UC, Perki- chloramphenicol. Pre-cultures were set up from colonies to grow overnight in 50 nElmer) was used per radioactive reaction. Reactions were stopped by adding ml LB at 37 °C. Culture volumes were increased to 500 ml and grown until OD 0.5. SDS-sample buffer. To detect radioactive signal, samples transferred to PVDF 3 Temperature was then reduced to 18 °C for induction with 0.1 mM IPTG membrane were dried 1 h, sprayed thrice with EN HANCE Spray (6NE970C, (800–050-IG, Wisent) overnight. 1 L cultures were spun down; pellets were PerkinEmer), while allowing drying for 30 min and rotating 90° following resuspended in binding buffer (50 mM NaPi pH 7, 500 mM NaCl, 10% Glycerol), each spray. Tritium signal was further enhanced with Transcreen LE − sonicated thrice 1 min each at power 7 and centrifuged at 27,216× g for 30 min, 4 ° (Z374253, Sigma-Aldrich) during the 2-week exposure at 80 °C. Alterna- 3 C. Soluble fractions were incubated with GST beads for 90 min at 4 °C. For p62, tively, radioactive polyacrylamide gels were soaked in EN HANCE liquid beads were then washed and eluted in 10 mM glutathione, 50 mM Tris pH 8.6, 150 (6NE9701, PerkinElmer) for 30 min at room temperature, washed six times in mM NaCl for 1 h at 4 °C. The eluate was dialyzed overnight with and without TEV water (5 min × 3 times and 15 min × 3 times), vacuum dried for 3 h at 60 °C − in 50 mM Tris pH 8, 150 mM NaCl, 0.5 mM EDTA, and 5 mM β-mercaptoethanol. and exposed for 2 weeks at 80 °C. Uncleaved GST-p62 (0.81 μg/μl) was frozen in 10% glycerol. Cleaved p62 was separated from GST by gel filtration and frozen in 10% glycerol (0.21 μg/μl). For SMN1, the beads were washed and eluted overnight in 1 M NaCl, 50 mM Tris pH 7 Nuclear/cytoplasmic subcellular fractionation at 4 °C to yield recombinant SMN1 (0.9 μg/μl). The protein was frozen in 10% . Cells washed with cold PBS were glycerol. lysed in hypotonic lysis buffer (10 mM HEPES pH 7.9, 10 mM KCl, 1.5 mM MgCl2, 0.5 mM DTT, protease inhibitors), incubated on ice for 10 min and homogenized with 15 strokes of a Dounce glass homogenizer. Supernatant was collected as the GST-pulldown assay with recombinant proteins. Recombinant p62 was mixed in cytoplasmic extract after centrifuging the lysate at 2000× g for 10 min. To collect buffer A (20 mM HEPES pH 7.9, 150 mM NaCl, 0.5 mM EDTA, 1 mM DTT, 10% the nuclear extract, the pellet was washed in hypotonic lysis buffer, further cen- glycerol, 0.1% triton-X-100) and pre-cleared with 5 μl of a 50% slurry of trifuged at 2000× g for 10 min, resuspended in NP-40 buffer (50 mM Tris pH 8, glutathione-agarose beads(16100, Pierce) and 5 μg of GST for 2 h at 4 °C with end- 400 mM NaCl, 5 mM EDTA pH 8, 1% NP-40, 0.2% SDS) and sonicated for 15 s at over-end mixing. Following centrifugation at 9000× g, 10 min, 4 °C, pre-cleared power 3. For western blotting, the fractions were run in equal volume proportions. supernatants were collected and incubated with 5 μl of a 50% slurry of glutathione- For quantification, cytoplasmic fractions were normalized to GAPDH while agarose beads and either 0.25 μg of GST or GST-C9ORF72 (H00203228-P01, nuclear fractions were normalized to methylated histone H3 on lysine 27—H3K27.

NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications 15 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7

Quantification of nuclear/cytoplasmic fluorescence. DAPI nuclear staining and those with MIDAS p-values 0.01. Additionally, the cumulative density of -log10 of brightfield images were used respectively to draw nuclear and cytoplasmic regions the best probeset MIDAS p-value per gene was plotted for all genes, as well as the 41 of interest on individual cells. Mean fluorescence intensity was measured for each set with higher FUS motif (GTGGT ) density than the mean (p-value 0.01) by a of the two regions per cell using Fiji and thereby represented as a ratio. one-sided proportion test.

Mass spectrometry. Chemicals used include urea, dithiothreitol (DTT), ammo- Analysis of putative interactomes of C9ORF72 and P62. Protein intensity values obtained from MaxQuant were used for the statistical confidence assessment nium bicarbonate (ABC), and iodoacetamide (IAA), all of which were purchased 62 from Sigma-Aldrich (St. Louis, MO). HPLC grade water and acetonitrile (ACN) of C9ORF72 and P62 interactions by SAINTq . For each protein, SAINTq were from J.T Baker. Formic acid (FA) and citric acid were obtained from Merck compares its intensity values in the biological replicates of the immunoprecipita- (Darmstadt, Germany). Trypsin was purchased from Promega (Madison, WI). All tions (IPs) against those in the IgG control replicates under the same experimental the chemicals were of analytical purity grade except ACN and FA, which were of conditions. SAINTq then assigns each protein a Bayesian False Discovery Rate fi HPLC grade. (BFDR) that is based on the level of signi cance of the intensity difference between Sliced protein bands were digested as reported previously59.Briefly, gel slices the immunoprecipitations and the controls. Proteins with a BFDR < 0.05 were fi were cut into smaller pieces, shrunk using 100% ACN, then reduced and alkylated considered as high-con dence protein-protein interactions. C9ORF72 and P62 interacting proteins were intersected with known stress granule proteins22 and sequentially. Trypsin solution of 10 ng/µL was used to cover the gel pieces for 63 digestion overnight. Tryptic peptides were then extracted using 80% ACN. The ALS-linked genes . Interactomes were also overlapped with published datasets of arginine dimethylated proteins and potential PRMT5 substrates44,64–67. Potential solution was then dried down and reconstituted in 20 µL 0.5% formic acid, and 65 4 µL was loaded for MS analysis. PRMT5 substrates from Supplementary Table 1 of which had SILAC H/L ratio > LCMS Analysis: Eksigent 2D + nanoLC system (Dublin, CA) was connected to 1.02 for at least two out of the three following treatment conditions: (i) EPZ015666 —selective PRMT5 inhibitor, (ii) siPRMT5_Exp1, and (iii) siPRMT5_Exp2 were a Q-Exactive mass spectrometer (Thermo Electron, Waltham, MA), equipped with 68 a nano-electrospray interface operated in positive ion mode. The solvent system included. Cytoscape 3.4.0 plugin BiNGO was used to perform GO term analyses. consists of buffer A of 0.1% FA in water, and buffer B of 0.1%FA in 80% Overrepresented categories were chosen after Benjamini & Hochberg False Dis- acetonitrile. Dried down protein digests were acidified with 0.5% (v/v) formic acid covery Rate correction. Hypergeometric statistical test was used to ascertain the p- and loaded on a 75 μm I.D. × 150 mm fused silica analytical column packed in- value of GO term enrichment. house with 1.9 μm ReproSil-Pur C18 beads (100 Å; Dr. Maisch GmbH, Ammerbuch, Germany) at a flow rate of 500 nL/min for 15 min. Then the flow rate Silver staining. Immunoprecipitations of P62, FUS, and C9ORF72 were per- was changed to 200 nL/min to perform the peptide separation. Gradient elution formed as described above except in a sterile environment to reduce keratin was set as 5–35% buffer B per hour. The spray voltage was set to 2.0 kV and the contamination. Following the final washes in NP-40 lysis buffer, magnetic beads temperature of the heated capillary was set to 300 °C. The instrument method were resuspended in 1% SDS (in lysis buffer), boiled at 99 °C for 5 min; then DTT consisted of one full MS scan from 300 to 1800 m/z followed by data-dependent was added to 10 mM and samples were boiled at 99 °C for 5 min prior to silver MS/MS scan of the 12 most intense ions, a dynamic exclusion repeat count of 1 in staining. Samples were run on precast protein gels (4561094, Bio-Rad) and pro- 30 s, and an exclusion duration of 30 s. The full mass was scanned in an Orbitrap cessed for silver staining as per manufacturer’s instructions (17-1150-01, GE analyzer with R = 70,000 (defined at m/z 400) for MS1 and 17,500 for MS2. To Healthcare). Protein lanes were then excised for trypsin digestion. improve the mass accuracy, all the measurements in the orbitrap mass analyzer were performed with a real time internal calibration by the lock mass of MTT Assay ™ background ion 445.120025. The charge state rejection function was enabled, and . Vybrant MTT Cell Proliferation Assay Kit (V13154, Thermo Fisher Scientific) was employed in accordance with manufacturer’s guidelines. charge states with unknown and single charge state were excluded for subsequent MS/MS analysis. All data were recorded with Xcalibur software (Thermo Fisher Scientific, San Jose, CA). Patient information, tissue collection, and consent. Patient tissues were col- Data Analysis: The peak lists of the raw files were processed and analyzed with lected with informed consent with approval from the Sunnybrook Health Sciences MaxQuant (Version 1.5.2.8)60 against UniProt human protein database, including Centre Research Ethics Board and the Ottawa Hospital Research Ethics Board. commonly observed contaminants. Cysteine carbamidomethylation was selected as Specimens were also provided by the Department of Veterans Affairs Bior- a fixed modification; methionine oxidation, protein N-terminal acetylation and epository. Overall, two non-ALS control cases, five ALS cases carrying C9ORF72 arginine methylation were set as variable modifications. Enzyme specificity was set repeat expansions, and three ALS cases not carrying C9ORF72 repeat expansions to trypsin, not allowing for cleavage N-terminal to proline. Up to two missing were used in the study. cleavages of trypsin were allowed. The precursor ion mass tolerances were 7 ppm, and fragment ion mass tolerance was 20 ppm. Razor and unique peptides were Statistical analysis. Statistical analyses were performed using two-tailed Student’s used for LFQ quantitation. FDR was set at 0.01 on protein, peptide and t-tests in Microsoft Excel. A p-value of <0.05 was considered statistically significant. modification of specific sites; a minimum length of seven amino acids was used for Significance was denoted as follows: *p-value < 0.05, **p-value < 0.01, ***p-value < peptide identification. For protein identification, if the identified peptide sequence 0.001. of one protein was equal to or contained another protein’s peptide set, these two proteins were grouped together by MaxQuant and reported as one protein group. Data availability. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the Bioinformatic analyses of splicing and FUS-motif enrichment. To analyze dataset identifiers PXD009759, PXD009741 and PXD009760. Dataset legends and potential splicing events using published microarray data from p62 knockout mice fi reviewer account details are included in the Supplementary Information. The the following MoGene-1_0-st-v1 chip les were used: (GSM1522557, GSM1522558 authors declare that all data supporting the findings of this study are available p62 knockout. GSM1522560 GSM1522562 wild type). The method outlined in61 fi fi within the article and its Supplementary Information les, or from the corre- with Affymetrix Power Tools 1.17.0 and MoGene-1_0-st-v1.r4 annotation les was sponding author upon reasonable request. used to identify splicing candidate genes. Genomic locations for probesets were taken from Affymetrix annotations: MoGene-1_0-st-v1.na35.mm10.probeset.csv. Genomic locations for exons and genes, as well as gene to probeset associations Received: 15 May 2017 Accepted: 21 June 2018 were downloaded from ENSEMBL BioMart mmusculus_gene_ensembl, dec2015. archive.ensembl.org. Genomic introns sequences were determined by selecting the genomic interval of each ENSEMBL gene with status “known” and removing intervals for exons associated with that gene. Genes were considered putatively spliced if they had MIDAS p-values < 0.01. Instances of the motif GTGGT iden- References fi 41 ti ed as the most frequent FUS-binding motif in intronic sequences were then 1. Mizushima, N. & Komatsu, M. Autophagy: renovation of cells and tissues. Cell counted and normalized to the total sequence length of mRNAs. 147, 728–741 (2011). The list of 957 mRNAs whose splicing was putatively affected by FUS was taken 2. Johansen, T. & Lamark, T. Selective autophagy mediated by autophagic from Table 6 in41 and 862 were found to be potentially mappable to the above adapter proteins. Autophagy 7, 279–296 (2011). derived genes based on MGI gene symbols in the ENSEMBL tcid annotations. 1481 3. Pankiv, S. et al. p62/SQSTM1 binds directly to Atg8/LC3 to facilitate putatively spliced MGI genes were identified in our analysis; there are 22,509 degradation of ubiquitinated protein aggregates by autophagy. J. Biol. Chem. unique MGI genes in the microarray gene expression set. Ninety-seven of the genes 282, 24131–24145 (2007). whose splicing was putatively affected by FUS41 were also found among the 1481 4. Khaminets, A., Behl, C. & Dikic, I. Ubiquitin-dependent and independent putatively spliced genes in p62 knockout mice. The predicted probability – distribution was generated using the R dhyper function and the above parameters. signals in selective autophagy. Trends Cell Biol. 26,6 16 (2016). The probability of seeing 97 or greater overlapping gene symbols by chance is 5. Heydrick, S. J., Lardeux, B. R. & Mortimore, G. E. Uptake and degradation of 1.57e-07. The cumulative fraction of -log10 of the best probeset MIDAS p-value per cytoplasmic RNA by hepatic lysosomes. Quantitative relationship to RNA – gene was plotted for all genes, as well as for the gene set matching FUS targets and turnover. J. Biol. Chem. 266, 8790 8796 (1991).

16 NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7 ARTICLE

6. Guo, H. et al. Autophagy supports genomic stability by degrading 36. Boisvert, F. M., Cote, J., Boulanger, M. C. & Richard, S. A proteomic analysis retrotransposon RNA. Nat. Commun. 5, 5276 (2014). of arginine-methylated protein complexes. Mol. Cell. Proteom. 2, 1319–1330 7. Buchan, J. R., Kolaitis, R. M., Taylor, J. P. & Parker, R. Eukaryotic stress (2003). granules are cleared by autophagy and Cdc48/VCP function. Cell 153, 37. Kurosawa, M. et al. Depletion of p62 reduces nuclear inclusions and 1461–1474 (2013). paradoxically ameliorates disease phenotypes in Huntington’s model mice. 8. Li, S., Yang, P., Tian, E. & Zhang, H. Arginine methylation modulates Hum. Mol. Genet. 24, 1092–1105 (2015). autophagic degradation of PGL granules in C. elegans. Mol. Cell 52, 421–433 38. Kim, B. et al. TUT7 controls the fate of precursor microRNAs by using three (2013). different uridylation mechanisms. EMBO J. 34, 1801–1815 (2015). 9. Protter, D. S. & Parker, R. Principles and properties of stress granules. Trends 39. Bussing, I., Slack, F. J. & Grosshans, H. let-7 microRNAs in development, stem Cell Biol. 26, 668–679 (2016). cells and cancer. Trends Mol. Med. 14, 400–409 (2008). 10. Panas, M. D., Ivanov, P. & Anderson, P. Mechanistic insights into mammalian 40. Emde, A. et al. Dysregulated miRNA biogenesis downstream of cellular stress stress granule dynamics. J. Cell. Biol. 215, 313–323 (2016). and ALS-causing mutations: a new mechanism for ALS. EMBO J. 34, 11. Lagier-Tourenne, C. & Cleveland, D. W. Rethinking ALS: the FUS about 2633–2651 (2015). TDP-43. Cell 136, 1001–1004 (2009). 41. Lagier-Tourenne, C. et al. Divergent roles of ALS-linked proteins FUS/TLS 12. Scekic-Zahirovic, J. et al. Toxic gain of function from mutant FUS protein is and TDP-43 intersect in processing long pre-mRNAs. Nat. Neurosci. 15, crucial to trigger cell autonomous motor neuron loss. EMBO J. 35, 1077–1097 1488–1497 (2012). (2016). 42. Matsumoto, G., Wada, K., Okuno, M., Kurosawa, M. & Nukina, N. Serine 403 13. Li, Y. R., King, O. D., Shorter, J. & Gitler, A. D. Stress granules as crucibles of phosphorylation of p62/SQSTM1 regulates selective autophagic clearance of ALS pathogenesis. J. Cell. Biol. 201, 361–372 (2013). ubiquitinated proteins. Mol. Cell 44, 279–289 (2011). 14. Teyssou, E. et al. Mutations in SQSTM1 encoding p62 in amyotrophic lateral 43. Zhang, Y. et al. SEPA-1 mediates the specific recognition and degradation of P sclerosis: genetics and neuropathology. Acta Neuropathol. 125, 511–522 granule components by autophagy in C. elegans. Cell 136, 308–321 (2009). (2013). 44. Guo, A. et al. Immunoaffinity enrichment and mass spectrometry analysis of 15. Renton, A. E. et al. A hexanucleotide repeat expansion in C9ORF72 is the protein methylation. Mol. Cell. Proteom. 13, 372–387 (2014). cause of chromosome 9p21-linked ALS-FTD. Neuron 72, 257–268 (2011). 45. Dormann, D. et al. Arginine methylation next to the PY-NLS modulates 16. DeJesus-Hernandez, M. et al. Expanded GGGGCC hexanucleotide repeat in Transportin binding and nuclear import of FUS. EMBO J. 31, 4258–4275 noncoding region of C9ORF72 causes chromosome 9p-linked FTD and ALS. (2012). Neuron 72, 245–256 (2011). 46. Dolzhanskaya, N., Merz, G., Aletta, J. M. & Denman, R. B. Methylation 17. Haeusler, A. R., Donnelly, C. J. & Rothstein, J. D. The expanding biology of regulates the intracellular protein-protein and protein-RNA interactions of the C9orf72 nucleotide repeat expansion in neurodegenerative disease. Nat. FMRP. J. Cell. Sci. 119, 1933–1946 (2006). Rev. Neurosci. 17, 383–395 (2016). 47. Maharjan, N., Kunzli, C., Buthey, K. & Saxena, S. C9ORF72 regulates stress 18. Waite, A. J. et al. Reduced C9orf72 protein levels in frontal cortex of granule formation and its deficiency impairs stress granule assembly, amyotrophic lateral sclerosis and frontotemporal degeneration brain with the hypersensitizing cells to stress. Mol. Neurobiol. 54(4), 3062–3077 (2017). C9ORF72 hexanucleotide repeat expansion. Neurobiol. Aging 35, e1775–e1779 48. Webster, C. P. et al. The C9orf72 protein interacts with Rab1a and the ULK1 (2014). 1779e1713. complex to regulate initiation of autophagy. EMBO J. 35, 1656–1676 (2016). 19. Xiao, S. et al. Isoform-specific antibodies reveal distinct subcellular 49. Brandman, O. & Hegde, R. S. Ribosome-associated protein quality control. localizations of C9orf72 in amyotrophic lateral sclerosis. Ann. Neurol. 78, Nat. Struct. Mol. Biol. 23,7–15 (2016). 568–583 (2015). 50. Lykke-Andersen, S. & Jensen, T. H. Nonsense-mediated mRNA decay: an 20. Cooper-Knock, J. et al. Sequestration of multiple RNA recognition motif- intricate machinery that shapes transcriptomes. Nat. Rev. Mol. Cell Biol. 16, containing proteins by C9orf72 repeat expansions. Brain 137, 2040–2051 665–677 (2015). (2014). 51. Chew, J. et al. Neurodegeneration. C9ORF72 repeat expansions in mice cause 21. Farg, M. A. et al. C9ORF72, implicated in amytrophic lateral sclerosis and TDP-43 pathology, neuronal loss, and behavioral deficits. Science 348, frontotemporal dementia, regulates endosomal trafficking. Hum. Mol. Genet. 1151–1154 (2015). 23, 3579–3595 (2014). 52. Mizielinska, S. et al. C9orf72 repeat expansions cause neurodegeneration in 22. Jain, S. et al. ATPase-modulated stress granules contain a diverse proteome Drosophila through arginine-rich proteins. Science 345, 1192–1194 (2014). and substructure. Cell 164, 487–498 (2016). 53. Peters, O. M. et al. Human C9ORF72 hexanucleotide expansion reproduces 23. Turinsky, A. L., Razick, S., Turner, B., Donaldson, I. M. & Wodak, S. J. RNA foci and dipeptide repeat proteins but not neurodegeneration in BAC Navigating the global protein-protein interaction landscape using iRefWeb. transgenic mice. Neuron 88, 902–909 (2015). Methods Mol. Biol. 1091, 315–331 (2014). 54. Komatsu, M. et al. Homeostatic levels of p62 control cytoplasmic inclusion 24. Geisler, S. et al. PINK1/Parkin-mediated mitophagy is dependent on VDAC1 body formation in autophagy-deficient mice. Cell 131, 1149–1163 (2007). and p62/SQSTM1. Nat. Cell Biol. 12, 119–131 (2010). 55. Gibbings, D. et al. Selective autophagy degrades DICER and AGO2 and 25. Wong, Y. C. & Holzbaur, E. L. Optineurin is an autophagy receptor for regulates miRNA activity. Nat. Cell Biol. 14, 1314–1321 (2012). damaged mitochondria in parkin-mediated mitophagy that is disrupted by an 56. Gibbings, D. et al. Human prion protein binds Argonaute and promotes ALS-linked mutation. Proc. Natl Acad. Sci. USA 111, E4439–E4448 (2014). accumulation of microRNA effector complexes. Nat. Struct. Mol. Biol. 19, 26. Souquere, S. et al. Unravelling the ultrastructure of stress granules and 517–524 (2012). S511. associated P-bodies in human cells. J. Cell. Sci. 122, 3619–3626 (2009). 57. Rizk, A. et al. Segmentation and quantification of subcellular structures in 27. Deosaran, E. et al. NBR1 acts as an autophagy receptor for peroxisomes. J. fluorescence microscopy images using Squassh. Nat. Protoc. 9, 586–596 Cell. Sci. 126, 939–952 (2013). (2014). 28. Martinez-Vicente, M. et al. Cargo recognition failure is responsible for 58. Sheffield, P., Garrard, S. & Derewenda, Z. Overcoming expression and inefficient autophagy in Huntington’s disease. Nat. Neurosci. 13, 567–576 purification problems of RhoGDI using a family of “parallel” expression (2010). vectors. Protein Expr. Purif. 15,34–39 (1999). 29. Yamano, K., Fogel, A. I., Wang, C., van der Bliek, A. M. & Youle, R. J. 59. Shevchenko, A., Tomas, H., Havlis, J., Olsen, J. V. & Mann, M. In-gel digestion Mitochondrial Rab GAPs govern autophagosome biogenesis during for mass spectrometric characterization of proteins and proteomes. Nat. mitophagy. eLife 3, e01612 (2014). Protoc. 1, 2856–2860 (2006). 30. Sahu, R. et al. Microautophagy of cytosolic proteins by late endosomes. Dev. 60. Cox, J. & Mann, M. MaxQuant enables high peptide identification rates, Cell. 20, 131–139 (2011). individualized p.p.b.-range mass accuracies and proteome-wide protein 31. Tradewell, M. L. et al. Arginine methylation by PRMT1 regulates nuclear- quantification. Nat. Biotechnol. 26, 1367–1372 (2008). cytoplasmic localization and toxicity of FUS/TLS harbouring ALS-linked 61. Lockstone, H. E. Exon array data analysis using Affymetrix power tools and R mutations. Hum. Mol. Genet. 21, 136–149 (2012). statistical software. Brief. Bioinform. 12, 634–644 (2011). 32. Bedford, M. T. & Clarke, S. G. Protein arginine methylation in mammals: 62. Teo, G. et al. SAINTq: scoring protein-protein interactions in affinity who, what, and why. Mol. Cell 33,1–13 (2009). purification—mass spectrometry experiments with fragment or peptide 33. Cote, J. & Richard, S. Tudor domains bind symmetrical dimethylated intensity data. Proteomics 16, 2238–2245 (2016). arginines. J. Biol. Chem. 280, 28476–28483 (2005). 63. Abel, O., Powell, J. F., Andersen, P. M. & Al-Chalabi, A. ALSoD: a user- 34. Hua, Y. & Zhou, J. Survival motor neuron protein facilitates assembly of stress friendly online bioinformatics tool for amyotrophic lateral sclerosis genetics. granules. FEBS Lett. 572,69–74 (2004). Hum. Mutat. 33, 1345–1351 (2012). 35. Sun, S. et al. ALS-causative mutations in FUS/TLS confer gain and loss of 64. Bremang, M. et al. Mass spectrometry-based identification and function by altered association with SMN and U1-snRNP. Nat. Commun. 6, characterisation of lysine and arginine methylation in the human proteome. 6171 (2015). Mol. Biosyst. 9, 2231–2247 (2013).

NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications 17 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-05273-7

65. Larsen, S. C. et al. Proteome-wide analysis of arginine monomethylation of FUS and p62 splicing datasets. G.Paris generated the FUS mutants, shRNA lentiviruses reveals widespread occurrence in human cells. Sci. Signal. 9, rs9 (2016). and stable cell lines with shRNA knockdowns. C.J. performed pulldowns with recom- 66. Uhlmann, T. et al. A method for large-scale identiﬁcation of protein arginine binant SMN, p62, and symmetrically dimethylated peptides. V.T. and J.F.C. generated methylation. Mol. Cell. Proteom. 11, 1489–1499 (2012). puriﬁed recombinant SMN and p62 and advised on pulldown assays using them. A.S. 67. Sylvestersen, K. B., Horn, H., Jungmichel, S., Jensen, L. J. & Nielsen, M. L. prepared and cultured primary mouse motor neurons. M.C. performed all of the Proteomic analysis of arginine methylation sites in human cells reveals remaining experiments. dynamic regulation during transcriptional arrest. Mol. Cell. Proteom.13, 2072–2088 (2014). Additional information 68. Shannon, P. et al. Cytoscape: a software environment for integrated models of – Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- biomolecular interaction networks. Genome Res. 13, 2498 2504 (2003). 018-05273-7.

Acknowledgements Competing interests: The authors declare no competing interests. We thank My Tran Trung and Charles Campbell for help with mouse work. We thank Reprints and permission information is available online at http://npg.nature.com/ −/− Herbert W. Virgin for p62 mice. Research in the D.G. group is supported by an open reprintsandpermissions/ operating grant from Canadian Institutes of Health Research (#142174). E.R. was supported by the Canadian Consortium on Neurodegeneration in Aging. M.L.A. is sup- Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in ported by a Discovery Grant from the National Science and Research Council of Canada published maps and institutional afﬁliations. and startup funds from the University of Ottawa. J.C. was supported by grants from Canadian Institutes of Health Research (MOP86746) and CureSMA Canada. J.W. was supported by Partners Investing in Parkinson’s Research group (Ottawa Hospital Foundation). J.R. was supported by the James Hunter and Family ALS Initiative and Open Access This article is licensed under a Creative Commons Brain Canada/ALS Canada. J.F.C. is supported by a grant from Canadian Institutes of Attribution 4.0 International License, which permits use, sharing, Health Research (MOP136816). M.C. was supported alternately by a 2015 ALS Society adaptation, distribution and reproduction in any medium or format, as long as you give Cycle of Hope Doctoral Research Award and a Vanier Canada Graduate Scholarship appropriate credit to the original author(s) and the source, provide a link to the Creative from the Canadian Institutes of Health Research. Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless Author contributions indicated otherwise in a credit line to the material. If material is not included in the ’ D.G. conceived the study. M.C. and D.G. designed and analyzed the experiments and article s Creative Commons license and your intended use is not permitted by statutory wrote the manuscript. J.C. helped design and analyze experiments particularly on argi- regulation or exceeds the permitted use, you will need to obtain permission directly from nine methylation and SMN. L.Z., J.K., E.R., and J.R. provided cerebellar tissues and the copyright holder. To view a copy of this license, visit http://creativecommons.org/ lumbar spinal cord sections of C9ORF72-ALS and sporadic-ALS patients. J.W. provided licenses/by/4.0/. cerebellar and hippocampal tissues and helped interpret the immunoﬂuorescence results of patient tissues. E.B.C. performed the splicing experiments. F.A.B. and M.L.A. performed the SAINT analysis. G.Palidwor designed and performed bioinformatics analysis © The Author(s) 2018

18 NATURE COMMUNICATIONS | (2018) 9:2794 | DOI: 10.1038/s41467-018-05273-7 | www.nature.com/naturecommunications